International Journal of Advances in Applied Sciences (IJAAS) Volume 9, issue 1, Mar. 2020 by ijaas_journal

ISSN: 2252-8814

IJAAS

International Journal of

Advances in Applied Sciences

International Journal of Advances in Applied Sciences (IJAAS) is a peer-reviewed and open access journal dedicated to publish significant research findings in the field of applied and theoretical sciences. The journal is designed to serve researchers, developers, professionals, graduate students and others interested in state-of-the art research activities in applied science areas, which cover topics including: chemistry, physics, materials, nanoscience and nanotechnology, mathematics, statistics, geology and earth sciences.

Editor-in-Chief: Qing Wang, National Institute of Advanced Industrial Science and Technology (AIST), Japan Co-Editor-in-Chief: Chen-Yuan Chen, National Pingtung University of Education, Taiwan, Province of China Bensafi Abd-El-Hamid, Abou Bekr Belkaid University of Tlemcen, Algeria Guangming Yao, Clarkson University, United States Habibolla Latifizadeh, Shiraz (SUTECH) University, Iran, Islamic Republic of EL Mahdi Ahmed Haroun, University of Bahri, Sudan

Published by:

Institute of Advanced Engineering and Science (IAES) Website: http://ijaas.iaescore.com Email: info@iaesjournal.com, ijaas@iaesjournal.com

Information for Authors International Journal of Advances in Applied Sciences (IJAAS) is an interdisciplinary journal that publishes material on all aspects of applied and theoretical sciences. The journal encompasses a variety of topics, including chemistry, physics, materials, nanoscience and nanotechnology, mathematics, statistics, geology and earth sciences. Submission of a manuscript implies that it contains original work and has not been published or submitted for publication elsewhere. It also implies the transfer of the copyright from the author to the publisher. Authors should include permission to reproduce any previously published material.

Ethics in publishing For information on Ethics in publishing and Ethical guidelines for journal publication (including the necessity to avoid plagiarism and duplicate publication) see: http://ijaas.iaescore.com/index.php/IJAAS/about/editorialPolicies#sectionPolicies

Paper Submission You must prepare and submit your papers as word document (DOC or DOCX). For more detailed instructions and IJAAS template please take a look and download at: http://ijaas.iaescore.com/index.php/IJAAS/about/submissions#onlineSubmissions The manuscript will be subjected to a full review procedure and the decision whether to accept it will be taken by the Editor based on the reviews. Manuscript must be submitted through our on-line system: http://ijaas.iaescore.com Once a manuscript has successfully been submitted via the online submission system authors may track the status of their manuscript using the online submission system.

IJAAS

International Journal of

Advances in Applied Sciences Vehicle accident management and control system using MQTT Sudha Senthilkumar, K. Brindha, Shashank Bhandari

1-11

Design and implementation of an effective web-based hybrid stemmer for Odia language Gouranga Charan Jena, Siddharth Swarup Rautaray

12-19

Simplified down sampling factor based modified SVPWM technique for cascaded inverter fed induction motor drive Kumar Bhukya, P. Satish Kumar

20-26

An efficient quantum multiverse optimization algorithm for solving optimization problems Samira Sarvari, Nor Fazlida Mohd. Sani, Zurina Mohd Hanapi, Mohd Taufik Abdullah

27-33

Trilateration based localization method using mobile anchor in wireless sensor networks M.G. Kavitha, Vinoth Kumar Kalimuthu, T. Jayasankar

34-42

An immune memory and negative selection to visualizing clinical pathways from electronic health record data Mouna Berquedich, Oulaid Kamach, Malek Masmoudi, Laurent Deshayes

43-50

A computer vision-based weed control system for low-land rice precision farming Olayemi Mikail Olaniyi, E. Daniya, J. G. Kolo, J. A. Bala, A. E. Olanrewaju

51-61

Intrusions detection using optimized support vector machine Mehdi Moukhafi, Khalid El Yassini, Bri Seddik

62-66

Top-K search scheme on encrypted data in cloud Katari Pushpa Rani, L. Lakshmi, Ch. Sabitha, B. Dhana Lakshmi, S. Sreeja

67-69

Software defined network emulation with OpenFLow protocol Tsehay Admassu Assegie

70-76

Responsibility of the contents rests upon the authors and not upon the publisher or editors

IJAAS

Vol. 9

No. 1

pp. 1-76

March 2020

ISSN 2252-8814

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 1~11 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp1-11



Vehicle accident management and control system using MQTT Sudha Senthilkumar, K. Brindha, Shashank Bhandari School of Information Technology and Engineering, Vellore Institute of Technology, India

Article Info

ABSTRACT

Article history:

The quick development of innovation has made our lives less demanding. Innovation has additionally expanded activity risks and street mishaps occur very often which cause tremendous death toll and damage to property on account of poor response from the people in charge of managing such incidents. The mishap recognition undertaking will give an ideal solution for this problem. An accelerometer or a Tilt Sensor can be used as part of an auto caution application with the goal that unsafe driving can be identified. It can be utilized as a crash or rollover finder of the vehicle amid and after a crash. With signals from a sensor, a serious situation because of an accident can be avoided or attended to at the earliest. At the point of time when a vehicle meets with an accident or an auto moves over, the tilt sensor recognizes the flag and promptly sends it to the microcontroller. The microcontroller sends the alarm message through the IoT Module including the location of the accident through the GPS Module to the police or control group by publishing it over the cloud. So, the crisis enable group can promptly follow the area through the GPS Module, subsequent to receiving and accepting the data. The area can likewise be seen on the Google maps. Vital move can be made if this data reaches the control group in time. This venture is valuable in recognizing the accident with the use of sensors. As a future execution, we can add a remote webcam to the current system in order to capture pictures of the scene of the accident.

Received May 8, 2019 Revised Aug 5, 2019 Accepted Oct 6, 2019 Keywords: GPS module IoT Vehicle tracking system

This is an open access article under the CC BY-SA license.

Corresponding Author: Sudha Senthilkumar, School of Information Technology and Engineering, Vellore Institute of Technology, Gorbachev Rd, Vellore, Tamil Nadu 632014, India. Email: sudha.s@vit.ac.in

INTRODUCTION The Internet of Things (IoT) is an arrangement of interrelated computing gadgets, mechanical and digital machines, objects, animals or individuals that are given one kind of an identifiers and the capacity to exchange information over a system without requiring human-to-human or human-to-PC communication. IOT is a new concept that has evolves from the convergence of wireless technologies. Wireless communication is the transfer of information or signal between two or more points that are not connected by an electrical conductor. IoT devices equipped with Wi-Fi allow the machine-to-machine communication. The sensor and actuator can be setup in different place but they are working together over an internet network [1]. Using IOT technique a vehicle tracking system (VTS) can be built. A vehicle tracking system combines the use of automatic vehicle location of individual vehicles with software that collects these fleet data for a comprehensive picture of vehicle locations. Modern vehicle tracking systems commonly use GPS technology for locating the vehicle, but other types of automatic vehicle location technology can also be used. Vehicle information can be viewed on electronic maps via internet with specialized software. Journal homepage: http://ijaas.iaescore.com



ISSN: 2252-8814

The history of vehicle tracking dates to the beginning of GPS technology in 1978. In the early years, the technology was not yet operational, due to an insufficient number of satellites orbiting the earth. On Jan 17, 1994, after years of gradual growth, the final of the first 24 satellites was launched, and the GPS system was considered fully operational. Early GPS was designed primarily only for military but in 1996, President Bill Clinton determined that the system would be an asset to civilians as well as the military. This policy change made GPS technology available to the average individual, including fleet managers, who could see the benefit of using the technology to keep tabs on their vehicles. In the early days of fleet tracking, in order to properly track a fleet, each vehicle had to be enabled with a costly GPS device. While helpful, these early systems were difficult to implement, costly to use and sometimes inconvenient for drivers and fleet management alike. Thus, it took several years for the concept to catch on. The modern fleet tracking system provides the necessary data to fleet managers allowing them to run their operations more efficiently. Reports on driver behavior, vehicle performance and fuel use all make it easier for the fleet manager to cut costs and increase efficiencies. These systems go beyond simple reporting of each vehicle’s location, offering fleet managers a wealth of information about their vehicles and their drivers. In many countries this VTS is available. But there is no system which can detect accident and also give the service of VTS. The proposed scheme considered to be a car safe project which can detect location of a car, and if there is any accident occur it can communicate automatically to the nearest police station, hospital and owner to reduce instant loss or damage. The quick development of innovation and framework has made our lives less demanding. The appearance of innovation has additionally expanded the activity risks and the street mishaps occur as often as possible which causes tremendous death toll and property on account of the poor crisis offices. 2.

OBJECTIVES The objective of this project is to detect an accident or a mishap at the earliest and inform the control groups in charge such as the police or hospitals so that they can react to this situation in minimum time possible and avoid any casualties and minimize damage to life and property by responding in a quick and efficient manner. This project is a combination of a VTS (Vehicle Tracking System) and finds out whether an accident has occurred or not and sends the message to the people in charge. The accident management data such as occurrence of the accident as well as the location of the accident are also published over a cloud and all the people that have access to the cloud server will have access to the data in order to provide the appropriate response measures. 2.1. Overview of VTS Vehicle tracking Systems are now widely used in day to day life of human beings. The fundamental concept of the Vehicle Tracking Systems is based on the GPS technology. Nowadays, most of the cars, buses, trucks, ships and other automotive vehicles are fitted with GPS trackers. Vehicle Tracking and Information System, Automatic Vehicle Locating System, Mobile Asset Management System are commonly referred to as Vehicle Tracking Systems. These are widely used in developed countries for a variety of reasons. GPS technology serves multiple purposes however, the most important being tracking of vehicles and for navigation purposed. Tracking systems enable a base or headquarter to monitor the location of the vehicle and keep track of the driver remotely without the intervention of the driver. Tracking systems were first developed for the shipping industry because they wanted to determine where each vehicle was at any given time. Passive systems were developed in the beginning to fulfil these requirements. For the applications which require real time location information of the vehicle these systems can’t be employed because they save the location information in the internal storage and location information can only be accessed when vehicle is available. To achieve Automatic vehicle location system that can transmit the location information in real time, active systems were developed. Real time vehicular tracking system incorporates a hardware device installed in the vehicle and a remote tracking server. The information is transmitted to the Tracking server using a Cloud protocol. This information is available to the users over the Internet. In active systems, users can get the real-time vehicle location on a Mobile Application with its current latitudinal and longitudinal data. This information is sent to the application through the MQTT protocol. 2.2. Accident identification and alerting The frequency of occurrence of accidents have increased exponentially over the past few years. Every year a large number of people suffer immense damages to themselves as well as others when they are involved in an accident. There is catastrophic collateral damage as well as many people end up dying in Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 1 – 11

Int J Adv Appl Sci

ISSN: 2252-8814



accidents. The main reason for this is that accidents are not identified effectively and the people that have been injured in the accident are not given the medical attention that they require. Accident alerting systems are basically systems that are able to correctly identify when the accident has occurred with the help of sensors that are present in the car at the time of the accident. Multiple sensors such as shock sensors, temperature sensors, tilt sensors etc can be used for the purpose of accident identification. Each of these sensors can be programmed with values that full under the acceptable and range and values that signify the occurrence of an accident. For eg, if the accident identification system is using a tile sensor, the tilt sensor could be having a value greater than 100. This means that the vehicle is currently in an upright and natural position and no accident has occurred. However, if the tilt sensor obtains a value of less than 100, it means that sensor has been moved from its original position and that an accident has occurred.Alerting systems are the extension of identification systems. Once an accident has occurred and the sensors have triggered an unacceptable value, this situation is then notified to the people in charge so that necessary action can be taken before there are fatal consequences. GSM has been widely used for the alerting portion as it is a very famous technology. However, we will be using a cloud protocol that sends a notification over the cloud to a mobile application which will intimate friends, family and control groups so that they can respond to the accident as soon as possible. 3. LITERATURE SURVEY 3.1. Previous design ideas Previously we had accident alert systems using GSM and GPS modem and Raspberry Pi. A piezoelectric sensor first senses the occurrence of an accident and gives its output to the microcontroller. The GPS detects the latitude and longitudinal position of a vehicle. The latitudes and longitude position of the vehicle is sent as message through the GSM. The static IP address of central emergency dispatch server is pre-saved in the EEPROM. Whenever an accident has occurred the position is detected and a message has been sent to the pre-saved static IP address (NodeMCU, 2018). Manasi Patil et al., described a better traffic management system using Raspberry pi and RFID technology. The vehicle has a raspberry pi controller fixed in it which is interfaced with sensors like gas sensor, temperature sensor and shock sensor. These sensors are fixed at a predetermined value before accident. When an accident occurs the value of one of the sensor changes and a message to a predefined number (of the ambulance) is sent through GSM. The GPS module which is also interfaced with the controller also sends the location of the vehicle. When the message is received by the ambulance, a clear route has to be provided to the ambulance. The ambulance has a controller ARM which is interfaced with the RFID tag sends electromagnetic waves. When an ambulance reaches the traffic signal the RFID reader which is placed on the joints detect the electromagnetic waves of the tag. If the traffic signal is red, then the readers goes through the database in fraction of seconds and turn the red-light green. And automatically in such condition the RFID on opposite joints turn the opposite signal red. This provides a clear route to the ambulance. Apurva Mane et al., described the methods for vehicle “collision detection and remote alarm device using Arduino. Key features of this design include real-time vehicle monitoring by sending its information regarding position (longitude, latitude), time, angle to the monitoring station and to the user/owners mobile that should help them to get medical help if accident or the theft occurs. Also, user/owner has an access to get real-time position of a vehicle in real time.” Whenever accident occurs MEMS and vibration sensor detects and sends the signals to microcontroller, by using GPS particular locations where accident has occurred is found, then GSM sends message to authorized members. Bhagya Lakshmi, proposed a FPGA Based Vehicle Tracking and Accident Warning system using GPS. FPGA is mainly used to track position of any vehicle and send automated message to pre-programmed number. The owner of vehicle, police to clear traffic, ambulance to save people can be informed by this device. FPGA controls and coordinate all parts used in system. With the help of accelerometer sensor, the exact position of the vehicle can be detected. It can also be predicted whether the vehicle is in normal position or upside down. The systems that have been mentioned above rely heavily on different types of sensors and different technologies that may not be readily available. Such systems have very high initial investments and have high maintenance and installation costs. Due to the high initial investments companies and people don’t opt for such technologies and prefer cheaper alternatives which are not as efficient as the systems that are mentioned above. However, matters such as these involving human lives and huge collateral damage need to be efficient and easy to maintain while being reliable. The proposed setup of the project does not involve the usage of multiple sensors, due to which the installation costs are relatively low. Also, the parts used in the proposed setup are readily available and replaceable in case of additional parts are required. The proposed setup of the project involves the usage of Tilt sensors or accelerometer to send a signal to the IoT device which includes the WiFi module. The GPS module that is attached allows us to track the location of the Vehicle accident management and control system using MQTT (Sudha Senthilkumar)



ISSN: 2252-8814

vehicle and provides the latitude and longitudinal data of the location of the accident. This data is then sent to the cloud using MQTT protocol and can be accessed by control groups or hospitals so that the people in charge can respond to the situation in minimum time and prevent loss of life and reduce the amount of damage caused. Figure 1 and Figure 2 show the block diagram and flowchart of the system. The connection of the component is shown in Figure 3.

Figure 1. Block diagram of system

Figure 2. Flowchart of the system

WIFI/IOT RXD

VT52, VT100, ANSI

TXD RTS Xmodem, Ymodem, Zmodem

CTS

TILT SENSOR

20 18

27.0 VOUT

30 31 32 1 2 9 10 11

19 22

PD0/RXD/PCINT16 PB0/ICP1/CLKO/PCINT0 PD1/TXD/PCINT17 PB1/OC1A/PCINT1 PD2/INT0/PCINT18 PB2/SS/OC1B/PCINT2 PD3/INT1/OC2B/PCINT19 PB3/MOSI/OC2A/PCINT3 PD4/T0/XCK/PCINT20 PB4/MISO/PCINT4 PD5/T1/OC0B/PCINT21 PB5/SCK/PCINT5 PD6/AIN0/OC0A/PCINT22 PB6/TOSC1/XTAL1/PCINT6 PD7/AIN1/PCINT23 PB7/TOSC2/XTAL2/PCINT7 AREF AVCC ADC6 ADC7

PC0/ADC0/PCINT8 PC1/ADC1/PCINT9 PC2/ADC2/PCINT10 PC3/ADC3/PCINT11 PC4/ADC4/SDA/PCINT12 PC5/ADC5/SCL/PCINT13 PC6/RESET/PCINT14

12 13 14 15 16 17 7 8 23 24 25 26 27 28 29

GPS RXD

VT52, VT100, ANSI

TXD RTS CTS

Xmodem, Ymodem, Zmodem

NODE MCU

Figure 3. Connection of the components 4. VARIOUS MODULE DESIGN 4.1. NodeMCU The NodeMCU is an open-source firmware and development kit that helps you to prototype your IoT product with few Lua script lines. The Development Kit based on ESP8266, integrates GPIO, PWM, IIC, 1-Wire and ADC all in one board. The ESP8266 is the name of a micro controller designed by Espressif Systems. The ESP8266 itself is a self-contained WiFi networking solution offering as a bridge from existing micro controller to WiFi and is also capable of running self-contained applications. This module comes with a built in USB connector and a rich assortment of pin-outs. With a micro USB cable, you can connect NodeMCU devkit to your laptop and flash it without any trouble, just like Arduino as shown in Figure 4. It is also immediately breadboard friendly. The MQTT library has been ported to the ESP8266 SoC platform and committed to the NodeMCU project. After this, the NodeMCU was able to support the MQTT IoT protocol using Lua to access the MQTT broker [1]. Figure 5 shows the pin diagram of NodeMCU. Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 1 – 11

Int J Adv Appl Sci

ISSN: 2252-8814



Figure 4. Schematic diagram of NodeMCU devkit

Figure 5. Pin diagram of NodeMCU 4.2. ESP8266 Wi-Fi unit The ESP8266 WiFi Module is a self-contained SOC with integrated TCP/IP protocol stack that can give any microcontroller access to your WiFi network. The ESP8266 is capable of either hosting an application or offloading all Wi-Fi networking functions from another application processor. Each ESP8266 module comes pre-programmed with an AT command set firmware, meaning, you can simply hook this up to Vehicle accident management and control system using MQTT (Sudha Senthilkumar)



ISSN: 2252-8814

your Arduino device and get about as much Wi-Fi-ability as a Wi-Fi Shield offers. The ESP8266 module is an extremely cost-effective board with a huge, and ever growing, community. This module has a powerful enough on-board processing and storage capability that allows it to be integrated with the sensors and other application specific devices through its GPIOs with minimal development up-front and minimal loading during runtime. Its high degree of on-chip integration allows for minimal external circuitry, including the front-end module, is designed to occupy minimal PCB area. The ESP8266 supports APSD for VoIP applications and Bluetooth co-existence interfaces, it contains a self-calibrated RF allowing it to work under all operating conditions, and requires no external RF parts [2]. It also has the capabilities of a microcontroller, very similar to Arduino and can be programmed using the Arduino IDE. Arduino IDE has an extension specifically for the ESP8266 so that it can be programmed with relative ease [2]. Figure 6 illustrates the ESP8266 pin diagram.

Figure 6. ESP8266 pin diagram 4.3. GPS module Global Positional System or GPS uses a Global Navigation Satellite System made up of a network of many satellites placed into the orbit by the US Dept. of Defense. A GPS Navigation Device or simply GPS is a device that is capable of receiving information from the GPS satellites in order to calculate the device’s geographical position. The GPS device receives this information from the satellites in terms of coordinates or latitudes and longitudes. Usage of appropriate software can allow us to plot those coordinates or latitudinal and longitudinal data on a map for easier and clearer understanding. Nowadays, most of the GPS modules are used in vehicles like cars, bikes, buses, trucks etc. Smartphones with GPS capability use (A-GPS) or Assisted GPS technology, which generally use the base station or towers to provide the tracking capability of the device. However, when we are outside the network range, the A-GPS will not be available for use as it requires the device to be in an area with network coverage. Since the GPS Module is connected with satellites, despite an area not having network connectivity, the GPS module will continue to function and provide the coordinates or the position of the system. “A GPS device can retrieve from the GPS system location and time information in all weather conditions, anywhere on or near the Earth. A GPS reception requires an unobstructed line of sight to four or more GPS satellites, and is subject to poor satellite signal conditions. In exceptionally poor signal conditions, for example in urban areas, satellite signals may exhibit multipath propagation where signals bounce off structures, or are weakened by meteorological conditions. Obstructed lines of sight may arise from a tree canopy or inside a structure, such as in a building, garage or tunnel [3]. The pin diagram of GPS module is shown in Figure 7.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 1 – 11

Int J Adv Appl Sci

ISSN: 2252-8814



Figure 7. Pin diagram of GPS module 4.4. Tilt sensor Acceleration is a measure of how quickly speed changes. Just as a speedometer is a meter that measures speed, an accelerometer is a meter that measures acceleration. We can use the accelerometer’s ability to sense acceleration to measure a variety of things that are useful to electronic and robotic projects and designs such as: Acceleration ,Tilt Angle, Incline, Rotation, Vibration ,Collision, Gravity [4]. For the purpose of our system we are using an accelerometer as our Tilt sensor which will be used to detect the occurrence of an accident or not.” 4.5. MQTT MQTT is a machine-to-machine/ “Internet of Things” publisher-subscriber based connectivity protocol. It was designed as an extremely lightweight publish/subscribe messaging transport. It is useful for connections with remote locations where a small code footprint is required and/or network bandwidth is extremely expensive. Message Queueing Telemetry Transport is an ISO standard publish-subscribe based messaging protocol. It works on top of the TCP/IP protocol. The publish-subscribe based messaging pattern requires a message broker. Message brokers are fundamentally programs that act as middleman between the sender and the receiver. It is the responsibility of the message broker to convert the message from the messaging protocol of the sender to the appropriate messaging protocol of the receiver. Message brokers are commonly used in computer networks or telecommunication where multiple software applications need to interact with one another by sending formally-defined messages.For the purpose of our system we are using the Mosquitto message broker which is lightweight and suitable for all devices from low power single board computers to full servers. The MQTT protocol provides a lightweight method of carrying out messaging using a publish/subscribe model. This makes it suitable for Internet of Things messaging such as with low power sensors or mobile devices such as phones, embedded computers or microcontrollers. The MQTT protocol is a good choice for wireless networks that experience varying levels of latency due to occasional bandwidth constraints or unreliable connections. Should the connection from a subscribing client to a broker get broken, the broker will buffer messages and push them out to the subscriber when it is back online. Should the connection from the publishing client to the broker be disconnected without notice, the broker can close the connection and send subscribers a cached message with instructions from the publisher [5]. 5. IOT, ALGORITHM ,TEST CASES 5.1. Internet of Things (IoT) The internet of things (IoT) is the network of physical devices, vehicles, buildings and other items embedded with electronics, software, sensors, actuators, and network connectivity that enable these objects to collect and exchange data. The IoT allows objects to be sensed and controlled remotely across existing network infrastructure, creating opportunities for more direct integration of the physical world into computerVehicle accident management and control system using MQTT (Sudha Senthilkumar)



ISSN: 2252-8814

based systems, and resulting in improved efficiency, accuracy and economic benefit. When IoT is augmented with sensors and actuators, the technology becomes an instance of the more general class of cyber-physical systems, which also encompasses technologies such as smart grids, smart homes, intelligent transportation and smart cities. Each thing is uniquely identifiable through its embedded computing system but is able to interoperate within the existing Internet infrastructure. Experts estimate that the IoT will consist of almost 50 billion objects by 2020 [6]. IoT will become an essential part of the lives of humans impacting us in a very big way. It will also become one of our sustenance requirements such as Internet, telephone, water, electricity. IoT’s impact will be similar to the impact of Internet at the time. However, Internet connected individual computers which could be at different locations whereas IoT will connect day-to-day usage objects which will have a strong impact in the physical world [6-8]. Data management is a crucial aspect in the Internet of Things. When considering a world of objects interconnected and constantly exchanging all types of information, the volume of the generated data and the processes involved in the handling of those data become critical. A long-term opportunity for wireless communications chip makers is the rise of Machine-to-Machine (M2M) computing, which one of the enabling technologies for Internet of Things [9, 10]. This technology spans abroad range of applications. While there is consensus that M2M is a promising pocket of growth, analyst estimates on the size of the opportunity diverge by a factor of four. Conservative estimates assume roughly 80 million to 90 million M2M units will be sold in 2014, whereas more optimistic projections forecast sales of 300 million units. Based on historical analyses of adoption curves for similar disruptive technologies, such as portable MP3 players and antilock braking systems for cars, it is believed that unit sales in M2M could rise by as much as a factor of ten over the next five years. There are many technologies and factors involved in the “data management” within the IoT context. Some of the most relevant concepts which enable us to understand the challenges and opportunities of data management are Data Collection and Analysis, Big Data, Semantic Sensor Networking, Virtual Sensors, Complex Event Processing [11-13]. 5.2. Algorithm Step 1: Start Step 2: Power on all modules Step 3: Wait for the tilt sensor to detect an accident Step 4: Once accident is detected, location of the accident is sent from the GPS Module to the NodeMCU Step 5: Send location as well as accident status to the cloud using MQTT protocol. Step 6: Rescue groups having access to the cloud are given real-time notification. Step 7: End 5.3. Test cases In order to check if the system is working correctly, we can perform a few test cases. If the system passes the test cases, we can safely say that the system is working as per the requirements. In order to check if individual components of the system are working we can perform the following the checks: Test case 1: Tilt Sensor: When the tilt sensor is in a natural or acceptable position the led light on the sensor does not glow. However as soon as it’s been tilted to a position which is unnatural or unacceptable, the led light glows. Therefore, we can check whether the tilt sensor is working correctly by tilting it at different angles and seeing whether the led starts glowing at angles above the threshold value. Test case 2: GPS Module: The GPS Module is responsible for giving the latitudinal and longitudinal values based on the position of the current system. We can check whether the GPS Module is working correctly if we are able to see the latitudinal and longitudinal values and checking it those values correspond to our current position. If the values are being displayed as 0,0 for latitude and longitude, it means that the GPS Module is not functioning as desired. Test case 3: MQTT App: As soon as the Tilt sensor detects the change in angle of inclination, it sends the notification to the MQTT application in real-time. If we are unable to see the values and the notification on the application, it means that the MQTT protocol is not working properly or the system is not connected to the server. If the values are displayed properly with the notification, it means that the system is connected to the server successfully and values are being displayed for the use of the user.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 1 – 11

Int J Adv Appl Sci



ISSN: 2252-8814

5.4. Implementation results After completing the test cases and making sure all the system components are fully functioning, we can start using the system. Once the system is implemented the following observations are made: a. Once the system is in an unnatural position, the led light on the tilt sensor starts glowing which signifies that the tilt sensor is in the position when an accident has occurred. The tilt sensor is used to detect the change in the g-forces in the object. The change in g-forces indicates the status of collision or crash of the object. Accelerometers are used for measuring vehicle body movement in various systems such as advanced braking system, passive restraint and electronically controlled suspension systems, etc. The tilt sensor has a more sensitivity and high stable in automotive environment. The tilt sensor module is essential module for vehicle accident detection system. It is used as automatic ECS (Emergency Calling System, which senses the change in the g-forces of the system. Actual thresholds g-forces for accident detection [14] are shown in Table 1. Table 1. Thresholds g-forces for accident detection [11] Accident harshness No Accident Mild Accident Medium Accident Severe Accident

Actual Maximum G Range representation 0-4 g 4-20 g 20-40g More than 40g

Due to the impact of the earth’s gravitational forces, change in the vehicle attitude will result in normal accelerations. The sensitivity of the tilt sensor is measured in the terms of the g-forces which are measured at the time of crash. The accident is detected by presence of vibration in the vehicle at the time of crash and the threshold value of the sensitivity can be measured. The g-forces of vehicles during crash are shown in Table 2. The location of the accident is intimated along with integration of google Map is added advantage in the proposed system which is not exist in the previous systems which is discussed [14, 15]. Table 2. Vehicle G-forces during crash S. No 1 2 3 4 5 6 7

g-forces 4g 10g 14g 40g 60g 70g 73g

Severity of Accident No Accident Mild Accident Mild Accident Major Accident Major Accident Major Accident Major Accident

Tilt sensor Sensor off Sensor off Sensor off Sensor Glow Sensor Glow Sensor Glow Sensor Glow

Message Activated No No No App Message Activated, Location is given App Message Activated, location is given. App Message Activated, location is given. App message Activated, location is given.

b. The MQTT Application displays 2 variables, one displays a notification whether an accident has occurred or not as shown in Figure 8 and Figure 9 respectively. “ACCIDENT OCCURRED” is displayed when the tilt sensor is tilted at an angle that is beyond the set threshold. “NO ACCIDENT” is displayed when the tilt sensor is in its natural position as shown in Figure 10.

Vehicle accident management and control system using MQTT (Sudha Senthilkumar)



ISSN: 2252-8814

Figure 8. Output of MQTT app when tilt sensor is in natural position

Figure 9. Output of MQTT app when an accident has occurred

Figure 10. Tilt sensor led glow 6.

CONCLUSION AND FUTURE SCOPE Hence the automatic alarm device for vehicle accidents has been implemented using NODE MCU microcontroller. This design is a system which can detect accidents in significantly less time and sends the basic information to first aid center within a few seconds covering geographical coordinates, the time in which a vehicle accident has occurred. Although the system is fully functional, some enhancements can be made in and around the system in order to make it even better and overcome some of the drawbacks of the existing system. A camera can be attached to the system in the vehicle so that the camera can click pictures at the time of the accident which may help the doctors in identifying the extent of injury of the people in the car and give the appropriate treatment. Also, the pictures will help the police in understanding the sequence of events and take action against the people responsible for the accident. A switch can be added to prevent false alarms or if the accident is not as severe and the driver and others are able to take care of themselves without needing urgent and immediate support. Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 1 – 11

Int J Adv Appl Sci

ISSN: 2252-8814



REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]

Apurva Mane and Jaideep Rana, “Vehicle Collision detection and Remote Alarm Device using Arduino,” International Journal of Current Engineering and Technology, Vol. 4, No. 3, June 2014. ESP8266 (n.d.). Retrieved March 27, 2018, [Online]. Available: https://en.wikipedia.org/wiki/ESP8266 GPS Module. Retrieved March 27, 2018, [Online]. Available: https://en.wikipedia.org/wiki/ GPS_navigation_device Kiran Sawant, Imran Bhole, Prashant Kokane, Piraji Doiphode, and Yogesh Thorat, “Accident Alert and Vehicle Tracking System,” International Journal of Innovative Research in Computer and Communication Engineering, Vol. 4, No. 5, May 2016. MQTT. Retrieved March 27, 2018, [Online]. Available: https:// en.wikipedia.org/wiki/MQTT Internet of Things. Retrieved March 27, 2018, [Online]. Available: https://en.wikipedia.org/wiki/Internet_of_things William M. Evanco, “The Impact of Rapid Incident Detection on Freeway Accident Fatalities,” MS – IVHS, 1996. Statistics of Accidents in India, 2008. [Online]. Available: http://www.nitawriter.wordpress.com/2008/ David A. Whitney and Joseph J Pisano TASC, Inc., Reading, Massachusetts, “Auto Alert: Automated Acoustic Detection of Incidents,” IDEA project, 1995. Manasi Patil, Aanchal Rawat, Prateek Singh, and Srishtie Dixit, “Accident Detection and Ambulance Control using Intelligent Traffic Control System,” International Journal of Engineering Trends and Technology (IJETT), Vol. 34, No. 8, April 2016. NodeMCU. Retrieved March 27, 2018, [Online]. Available: https://en.wikipedia.org/wiki/NodeMCU Sri Krishna Chaitanya Varma, Poornesh, Tarun Varma, and Harsha, “Automatic Vehicle Accident Detection and Messaging System Using GPS and GSM Modems,” International Journal of Scientific & Engineering Research, Vol. 4, No. 8, August 2013. V.Sagar Reddy, L. Padma Sree, and V. Naveen Kumar, “Design and Development of accelerometer-based System for driver safety,” International Journal of Science, Engineering and Technology Research (IJSETR), Vol. 3, No. 12, December 2014. EEE Technical Committee for Sensor Technology, October 2009. The IEEE P1451.6 Project. [Online]. Available: http://grouper.ieee.org/groups/1451/6/index.htm D. Punetha, D. Kumar, and V. Mehta, “Design and realization of the Accelerometer based Transportation System (ATS),” International Journal of Computer Applications, Vol. 49, No. 15, 2012.

Vehicle accident management and control system using MQTT (Sudha Senthilkumar)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 12~19 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp12-19



Design and implementation of an effective web-based hybrid stemmer for Odia language Gouranga Charan Jena, Siddharth Swarup Rautaray School of Computer Engineering, KIIT Deemed to be University, India

Article Info

ABSTRACT

Article history:

Stemmer is used for reducing inflectional or derived word to its stem. This technique involves removing the suffix or prefix affixed in a word. It can be used for information retrieval system to refine the overall execution of the retrieval process. This process is not equivalent to morphological analysis. This process only finds the stem of a word. This technique decreases the number of terms in information retrieval system. There are various techniques exists for stemming. In this paper, a new web-based stemmer has been proposed named as “Mula” for Odia Language. It uses the Hybrid approach (i.e. combination of brute force and suffix removal approach) for Odia language. The new born stemmer is both computationally faster and domain independent. The results are favourable and indicate that the proposed stemmer can be used effectively in Odia Information Retrieval systems. This stemmer also handles the problem of over-stemming and under-stemming in some extend.

Received May 3, 2019 Revised Aug 20, 2019 Accepted Jan 5, 2020 Keywords: Brute force Derivational suffixes Inflectional words Information retrieval Mula Odia stemmer

This is an open access article under the CC BY-SA license.

Corresponding Author: Gouranga Charan Jena, School of Computer Engineering, KIIT Deemed to be University, Bhubaneswar, Odisha, India. Email: jenagouranga2000@gmail.com

INTRODUCTION Stemming is a technique for removing inflectional or derived word to its stem. This technique is used to remove the suffix or prefix affixed in a word. This process finds stem of a word. Stemmer is essential for information retrieval system to refine the performance of the system. Its technique is not equivalent to morphological analysis. Its primary objective is to decreases the number of terms in an information retrieval system. Stemming technique can be used in information retrieval to decrease as many related words to a common form that is not in base form. For example, the English word “Computation” has different inflections such as ‘Comput’, ‘Compute, ‘Computing’, ’Computes’ etc. In this case stemmer can be used to reduce derived words into its root or stem word. Many stemmers had been developed for different languages, which reduce a word to its root/stem form. It ultimately reduces the index file size in an information retrieval system. In this way we can improve recall (i.e. the number of documents retrieved in response to a query.) of an IR system by effectively using stemmer in the background. Since many derivational words are mapping into one word i.e. root or stem. It ultimately reduces the volume of the index files in the IR system. There are several types of stemming algorithms exists and it differs in respect to their performance and accuracy. There are various algorithms used to find a stem of a word i.e. (a) Brute-force algorithms: It uses a lookup table that contains derived words with their corresponding roots. To find the root/stem of a word, the table is queried to find a matching inflection word. If a matching inflection word is found, then corresponding root returned. Otherwise, it fails. (b) Suffix-stripping Approach: It does not depend on any Journal homepage: http://ijaas.iaescore.com

Int J Adv Appl Sci

ISSN: 2252-8814



lookup table that consists of derived or inflected words and their root word relations. It simply uses a set "rules" which drives the algorithm. It finds the root/stem of the given input word based on that rules. (c) Lemmatization algorithms: This technique also called as text normalization. In Lemmatization root word is called Lemma. The POS is first identified of that language and then an attempt will be made to find the stem. The stemming rules will change based on a word's POS of that language. Lemmatization process ensures that the root or stem of the inflected words belongs to that language (d) Stochastic algorithms: It is based on probability method to detect the root form of a word. This trained on a table of root words to inflected words relations to develop a probabilistic model. (d) Affix Removal Approach: The name clearly suggests this approach is related to removing the suffixes or prefixes of a word. An Affix may be a prefix or a suffix. It comes under truncating method of the stemming algorithm. We found affixes are connected with nouns in Odia language. One can opt any of the above technique while designing stemmer. (e) Hybrid approach: This technique combines more than two methods as discussed above. It may merge the rule-based technique along with the probability method. (f) N-Gram Modeling: Many stemming methods used in the ngram technique of a word to select the correct stem for a word. Stemming plays a vital role to handle the vocabulary mismatch problem of an IR system. In this said problem, the query words mismatch with the document words. For example, when a user input a query word and the word does not exist in the vocabulary of the document then it may cause unreliable result. To avoid this problem, we have developed a new web-based hybrid stemmer using brute force with enhanced suffix stripping algorithm union that can be adopt in the Odia information retrieval system. The new stemmer is both computationally faster as well as domain-independent. 2.

RELATED WORKS In the study of information retrieval, researchers find stemming plays an important role. Stemming is not a new concept. Stemming techniques had invented since 1968. The first stemming algorithm was designed by Julie Beth Lovins [1]. After that many researchers continued investigating various approaches to this area of study and proposed several algorithms to improve its performance. Another stemmer in English was written by Martin Porter [2] in the year 1980. As compared to European languages as well as English, a few researches have been discovered in Indian Language. A Hindi stemmer [3] was proposed by Rao, Durgesh et al. based on suffix striping approach. A Bengali Morphological analyzer [4] was developed by Dasgupta et al. based on suffix striping approach. Stemming is the process by which the user inputs an inflected word to the trained model and the model produces the root/stem word according to its rule set. In this Paper we have developed a Stemmer based on Hybrid Approach. 3.

LITERATURE SURVEY ON ODIA STEMMER These are the few papers published on Odia Stemmer. Table 1 describes the paper details with key findings. Table 1. Literature survey on Odia stemmer Reference Sampa et al. [5]

Balbantray, R.C. et al. [6] Balabantaray, R.C. et al. [7] Sethi, Dhabal Prasad [8]

Key Findings Published a paper on Stemmer for Odia language. They used the suffix stripping approach to remove the inflectional suffixes. The limitation of this algorithm was it only predicts 88% accuracy. Published a paper FIRE 2012 Submission: MET Track Odia. They had used the affix removal algorithm. The system reads input text files from the folder. Firstly, it removes stop words from the input files against the stop word dictionary then matched the token with the root word dictionary. After that the input matched the suffix dictionary then removes the suffix and match with the root word. If the root word found then there is no further processing required. Presented a paper on Odia Text Summarization. Published a paper on Lightweight Stemmer for Odia Derivational Suffix. He used suffix stripping method to find the stems.

4. ODIA DERIVATIONAL MORPHOLOGY 4.1. Odia morphology The formal variants of a morpheme are called allomorphs of that morpheme. The variant may be phonologically or morphologically conditioned. A morpheme may be a free or a bound form. Alternatively, we can say that a word consists of one or more than one morpheme. From the point of view of its internal structure, a word may consist of (i) a root morpheme only (ii) a root and one or more non root morpheme or (iii) more than one root morpheme. The non-root morphemes are bound forms and are generally referred to as affixes. Roots enter into further morphological constructions and form a base while non-roots do not [9]. Design and implementation of an effective web-based hybrid stemmer for Odia … (Gouranga Charan Jena)



ISSN: 2252-8814

4.2. Word formation Word formation is concerned with those words which comprise more than one meaningful component called morphemes. The common morphological processes, which are involved in word formation, are inflection, derivation, reduplication, echo formation and contraction. The word formation process is shown in Figure 1. 4.2.1. Inflection Inflection is a morphological process by which words are formed with the help of bound forms, which are called inflectional affixes. Inflected words belong to the same form-class to which the root word belongs. 4.2.2. Derivation Derivation is a morphological process, which is concerned with the structure of the stems. In other words, word stems are formed by derivation. Two types of this process are generally distinguished and they are compounding and derivation. Compounding is a derivational process in which a stem is formed with two roots, the resultant stem belonging to the form class of at least one of the constituent roots. Derivation is a process of word formation in which a stem is formed with two roots or a root and an affix and the resultant stem does not belong to the form class of any of the constituents. Both inflectional and derivational affixes are involved in affixation. Depending on their position of occurrence with respect to the root, the affixes are classified into prefixes, suffixes and infixes. Prefixes precede the root, suffixes follow it and infixes occur within the root. 4.2.3. Reduplication Laurel J. Brinton in his structure of English: A Linguistic Introduction defines “Reduplication is a process similar to derivation, in which the initial syllable or the entire word is doubled, exactly or with a slight morphological change.” Reduplication is another morphological process in which a part of a root or the root itself is added to the root. This type of word formations is popular in Odia language. 4.2.3. Echo formation The partial repetition of a phoneme or syllable of the base may be called an echo-formation. In other words, if the initial phoneme/syllable of the base is replaced by another phoneme or syllable it has neither any individual occurrence nor any meaning of its own. It may be called as echo-formation. 4.2.4. Contraction Contraction is a process of word formation in which a syllable is dropped from the root. In Odia words are formed using different morphological process viz., inflection, compounding derivation, affixation, reduplication and contraction. Both prefixes and suffixes occur in Odia. The prefixes are used to form derived adjectives, verbal noun, agent noun, collective and reciprocals. The suffixes denote gender, number, case, tense, aspect, and mood.

Inflection

Contraction Derivation

Word Formation

Echo Formation

Reduplication

Figure 1. Word formation process

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 12 – 19

Int J Adv Appl Sci

ISSN: 2252-8814



The formal variants of a morpheme are called allomorphs of that morpheme. The variant may be phonologically or morphologically conditioned. A morpheme may be a free or a bound form. Alternatively, we can say that a word consists of one or more than one morpheme. From the point of view of its internal structure, a word may consist of (i) a root morpheme only; (ii) a root and one or more non root morpheme or; (iii) more than one root morpheme. The non-root morphemes are bound forms and are generally referred to as affixes. Roots enter into further morphological constructions and form a base while non-roots do not [9]. Odia morphology deals with the analysis, identification and description of structure of morpheme. Morphology deals with the structure of words. The basic unit is the focus of study in morphology is morpheme. For example: The word ବାଳକମାେନ the morphemes are ବାଳକ, ମାେନ. Morpheme is not always conveying a meaningful word in Odia. Any morpheme in Odia should be a root word, prefix or suffix. Morphemes are divided into five categories shown in Figure 2. The morpheme which are independent called free morpheme. Those morphemes are standalone in nature. It does not need to add with other to create a word. Examples of free and bound morpheme: ରାମ ଭାତ ଖାଉଛି । ରାମ ଭାତ(କୁ ) ଖାଉଛି| The morpheme ଭାତ is a stand-alone morpheme and morpheme (କୁ ) is a suffix. Most of the morphemes are bound type in Odia language.

Figure 2. Type of morpheme 4.3. Odia derivational morphology Derivational morphology deals with the addition of derivational suffixes with word stem to form word of different class (different part-of-speech). Like English, Odia derivational suffixes are added with root word to form different part-of-speech. They are in Table 2. Table 2. Detail description of Odia derivational morphology Categories

Examples Noun word + Derivational suffix = Adjective category

Noun to Adjective

ରୁପ+ଏଲି = ରୁେପଲି Adjective word + derivational suffix = noun words

Adjective to Noun

ଆଧୁନକ ି + ତା=ଆଧୁନକ ି ତା Adjective word + Suffix = Adjective word

Adjective to Adjective

ସରୁ+ଆ=ସରୁଆ Verbal word + Derivational suffix =Adjective word

Verb to Adjective

ପାନ+ଇଅ= ପାନିଅ Verbal word + Derivational suffix = Noun word

Verb to Noun

ବି +ଅଣା= ବି ଣା େଖଳ+ଆଳି=େଖଳାଳି

Design and implementation of an effective web-based hybrid stemmer for Odia … (Gouranga Charan Jena)



ISSN: 2252-8814

In derivational stemming, words that are derived, either by adding affixes to that stems or by performing changes at the morpheme boundary, are reduced to their stem form. Odia language has strong inflectional system can be classified as nominal inflection and verb inflection. Here we represent the rules using Panini Grammar. For example, ‘କଲମଗୁଡ଼କ ି ’ here stem କଲମ and suffix is ଗୁଡ଼କ ି . The details of nominal suffix and Verbal suffix are given in Table 3 and Table 4 [5]. In Odia we find some prefixes which is attached only on noun. There is 20 such type of prefixes in Odia. These are basically from Sanskrit. They are as shown in Table 5. Table 3. List of nominal suffixes in Odia ଏକ ବଚନ (Singular)

ବହୁ ବଚନ (Plural)

କାରକ (CaseRelationship)

ଏଜ,େଟ,ଟି,ଟା,ଟିଏ,ଟାଏ,◌୍ଞ

ଏଗୁଡ଼ଏ ି ,ଗୁଡ଼ାକ,ଗୁଡ଼କ ି ,ମାନ ,ମାେନ ,

କତା

ବିଭ ି Inflection ଥମା (1st Inflection) ତ ି ୀୟା (2nd Inflection) ତୃ ତୀୟା (3rd Inflection)

କୁ,େତ,ଟିକୁ,ଟାକୁ , ୁ ,କି, ୁ , ଠାକୁ ଠିକ,ି ଠିକୁ, େର

ି ୁ ,ମାନ ୁ , ୁ ଗୁଡ଼ାକୁ ,ଗୁଡ଼କ

ାରା,େଦଇ, ାରା,

ଚତୁ ଥୀ (4th Inflection)

କୁ, ,ଲାଗି,ପାଇଁ, ି ,କି, ୁ , ନିମେ , େଯାଗଁୁ , ନିମେ , ଲାଗି, ସକାେଶ ,ସକାେଶ

ପ ମୀ (5th Inflection)

ରୁଠଉଁ ,ଠଁୁ ,ଠାରୁ ,

ଷ ୀ (6th Inflection) ସ ମୀ (7th Inflection)

ର,ଠି,ଠାଇଁ ,ଏ , ର , , େଠଇଁଠ ,ତହ, େର,େଠଇଁ,ଠି,ଠାଇଁ,ଏ,ଠାେର, ତହଠ ,

କମ

ାରା ,ମାନ େଦଇ ,ମାନ ାରା , ମାନ େର ି ୁ ,ମାନ ୁ , ୁ ,ମାନ ଲାଗି ,ଗୁଡ଼ାକୁ,ଗୁଡ଼କ ନିମ ,ମାନ ନିମେ େତେଯାଗଁୁମାନ , ସକାେଶ ମାନ ଠାରୁ,ମାନ ଠଁୁ , ଠାରୁମାନ ଠୁ ,ମାନ ଠଉଁ , ମାନ

ଉପପଦ (NoncaseRelationship)

କରଣ ସ ଦାନ ଅପାଦାନ

ରି ,ମାନ ରି ,ମାନ ର ,

ସମ

ମାନ େର ,ମାନ ଠାଇଁ ,ମାନ ଠାେର, ମାନ େଠଇଁ ,ମାନ ଠ

ଅ କରଣ

Table 4. Odia verbal suffix କାଳ (Tense) ବ ମାନ କାଳ (Present Tense)

ଅତୀତ କାଳ (Past Tense)

ଭବିଷ ତ କାଳ (Future Tense)

ପୁରୁଷ (Person) ଥମ ପୁରୁଷ ତ ି ୀୟ ପୁରୁଷ ତୃ ତୀୟ ପୁରୁଷ ଥମ ପୁରୁଷ ତ ି ୀୟ ପୁରୁଷ ତୃ ତୀୟ ପୁରୁଷ ଥମ ପୁରୁଷ ତ ି ୀୟ ପୁରୁଷ ତୃ ତୀୟ ପୁରୁଷ

ଏକ ବଚନ (Singular Suffix)

ବହୁ ବଚନ (Plural Suffix)

ଉଅଛି,ଉଛି,ଉଥା ,ି ଏ ଉଅଛୁ , ଉଛୁ , ଉଥା ୁ , ର ଉଅଛି, ଉଛି, ଉଅଛ ,ି ଉଥା ା, ଉଥାେ , ଉଥାଆେ , ,ି ଉଆ ା ଇଲି,ଇଛି, ଇ ଲି, ଲି,ଇଥା ,ି ଇଅଛି ଇଲୁ ,ଇଛୁ , ଇ ଲୁ , ଉ ଲୁ , ଇଥା ୁ ଇଲା, ଇେଲ, ଇଛି, ଇଛ ,ି ଇ େଲ, ଇ ଲଲ, ଉ େଲ, ଇଥାଆ ା, ଇଥାଆେ ଇବି, ଇ ବି, ଥା ,ି ଉ ବି ଇବୁ , ଇ ବୁ , ଥାଆ ୁ , ଉ ବୁ ଇବ, ଇେବ, ଇ େବ, ଇ ବ, ଥାଆ , ଥାଆେ , ଉ ବ, ଉ େବ

ଉଛୁ , ଇଅଛୁ , ଉଥା ୁ , ଉ ଉଅଛ, ଉଛ, ଉଥା ଉଛ ି, ଉଅଛ ି, ଉଥାଆ ,ି ି ଗଲୁ , ଇ ଲୁ ,ଉ ଲୁ , ଇଥା ୁ ଇଲ, ଇଛ, ଇଅଛ, ଇ ଲ, ଉ ଲ, ଇଥା ଇେଲ, ଇଛ ,ି ଇ େଲ, ର େଲ, ଇଥାଆେ ଇବୁ , ଇ ବୁ , ଥାଆ ୁ , ଉ ବୁ , ଇବା ଇବ, ଇ ବ, ଥାଆ ୁ , ଉ ବ ଇେବ, ଇ େବ, ଥାଆେ , ଇ େବ

Table 5. List of Odia prefixes ଅ

ତି ଅଭି

ପରା

ଅପ

ସ

ସୁ ଅବ ଅତି

ନି ଅନୁ ଅପି

ଉ ଦୁ ଉପ

ନ ପରି ବି ଆ

PROPOSED METHODOLOGY FOR ODIA STEMMER We have proposed a new web-based stemmer based on hybrid approach (i.e. combination of brute force and suffix removal approach) [10] for Odia language. The proposed stemmer is both computationally inexpensive and domain independent. The algorithm of the proposed stemmer is described in Figure 3.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 12 – 19

Int J Adv Appl Sci

ISSN: 2252-8814



Figure 3. Flowchart of the algorithm Brute force search is called as exhaustive search. It searches all the possible solution from the data. Here it searches the root words present in the database. This technique uses a lookup table which contains inflected words and root words mapping. This technique [11] we create and store maximum possible inflected words along with their corresponding root word in a database table. When we give input to the system then brute force search is carry out and it inspects that whether the derivational words exist in the database. If the word is present in that table then it will give its corresponding stem or root word. If the word is not present in the table then it will go for suffix removal method to handle those words. Suffix removal is a rule-based approach in that certain rule set is defined. By applying those rule set suffixes are removed from the inflected or derived word, to find the stem/root. The new enhanced approach of suffix stripping algorithm. Figure 4 shows the stemmer user interface. Start Step 1: Enter derivational word that to be stemmed Step 2: The system removes the 3 characters suffixes, 2 character suffixes and 1 character suffixes from the derivational word if word length greater than three, and two respectively recursively. End The inflected word is processed by the stemmer in three steps. The steps are shown below. a. Input: The inflected Odia word/paragraph is entered as an input to the web-based system. Here “ଆଧୁନକ ି ାତା” is given as an Input word. b. Processing: Derivational/inflected word is searched by brute force method. It matches with the user searched word with the words exist in the database table. If the matching word is exist in the database then it will provide the stem of the word as output. If mismatch found then it searches for the alternate method called suffix stripping method i.e. the algorithm removes the suffixes recursively first 3 characters, then 2 characters and last 1 character with a condition that the inflected word must be greater than the suffix to find the stem/root of the word. c. Output Unit: In Output Unit, the result comes after the processing of word. The result after processing is “ଆଧୁନକ ି .” One Character derivational Suffix: ଅ, ଆ, ଇ, ଈ, ତ, ତି, ଯ, , ତ, ନ, ତା, , ଥା, ଦା, ଧା, କା, ରା Two Character derivational Suffix: ଅନ, ମୟ, ଅକ, ତର, ଉକ, ତମ, ଇନ, ଇ , ଇ ୁ , ଆଳୁ , ତବ , ଆଳ, ଇ , ଆଳି, ଉର, ଆରି, ସନ, ଇଆ, ଅଣ, ଉଆ, ଅଣା, ଉରା, ଅନି, ଏଲି, ଅ ା, କାର, ଅ ୀ, କୁ ଳା, ଆଣ, ବ , ଅଉ, ଚିଆ, ଉଆ, ପଣ, ଉଣି, ଖାନା, ଉଣା, େଖାର, ଏଣି, ଗର, ଇକ, ଗିରି, ଈୟ, ଆମି, ଈନ, ବାଜ, ଏୟ, ବାଲା, ଇତ, ଦାର, ଇଳ, ବିନ, ଇନ, ତନ, ବତ. Three Character derivational Suffix: ଅନୀୟ, ଉଆଳ, ଉଆଳି, ଶାଳିନ, ଈୟସ, ାନୀୟ, ଆଳିଆ. Design and implementation of an effective web-based hybrid stemmer for Odia … (Gouranga Charan Jena)



ISSN: 2252-8814

Figure 4. Stemmer user interface 6.

EVALUATION We have evaluated the stemmer by taking different set of words i.e. 100 words, 200 words, 300 words and so on to calculate the time taken to extract the root words as shown in Figure 5. We have not compared our Odia stemmer with any of the existing stemmer available for Odia language. Nowhere had we found the existing result to compare with the proposed stemmer. Table 6 shows the time to extract Odia root words. Table 6. Time taken to extract Odia root words Set No Set-1 Set-2 Set-3 Set-4 Set-5 Set-6 Set-7 Set-8 Set-9 Set-10

No of Words 100 200 300 400 500 600 700 800 900 1000

Time Taken (Sec.) 5.07 8.92 13.27 17.18 20.29 24.84 29.11 32.85 37.66 40.48

Figure 5. Evaluation graph

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 12 – 19

Int J Adv Appl Sci

ISSN: 2252-8814



7. APPLICATION OF STEMMER 7.1. Information retrieval Stemmer can be used in information retrieval [12] to reduce as many related words to a common form which is not in base form. 7.2. Indexing Stemmer can ultimately reduce the indexing size [13] of the documents and thus the retrieval process become faster. 7.3. Auto text summarization It reduces a text document to its summary. It can be used for text summarization [14]. 7.4. Cross-Language Information Retrieval (CLIR) Stemmer can be used in cross-language information retrieval [14] to reduce as many related words to a common form which is not in base form. Example suppose the user enter the query in English, it retrieves relevant document written in Odia. 8.

CONCLUSION This stemmer can be played a vital role for the performance of an Odia IR System. It can efficiently handle the problem of understemming and over stemming. In future we can extend this research by merging few other techniques by including some more data in to the database and also by using extra rules for suffix stripping approach. Then we can compare the results with this stemmer result. In this way we can conclude which merging technique is computationally faster. Accordingly, we pick the approach for building an Odia IR system. In this paper we designed a stemmer algorithm for Odia which removes derivational suffixes from derived word. This algorithm uses brute force approach and a new enhanced approach of simple suffix removal technique. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

J.B. Lovins, “Development of a stemming algorithm,” Mechanical Translation and Computational Linguistics, Vol. 11, No. 1 and 2, pp. 22-31, 1968. M.F. Porter, “An algorithm for suffix stripping,” Program, Vol. 14, No. 3, pp. 130-137, 1980. A. Ramanathan and Durgesh D. Rao, “A lightweight stemmer for Hindi,” 2003. S. Dasgupta and V. Ng, “Unsupervised morphological parsing of Bengali,” Lang. Resources & Evaluation, Vol. 40, pp. 311-330, 2006. S. Chaupattnaik, S.S. Nanda, and S. Mohanty, “A suffix stripping Algorithm for Odia Stemmer,” Int. Journal of Computational Linguistics and Natural Language Processing, Vol. 1, No. 1, 2012. R.C. Balabantaray, B. Sahoo, M. Swain, and D.K. Sahoo, “IIIT-BH FIRE 2012 Submission: MET Track Odia,” 2012. R.C. Balabantaray, B. Sahoo, M. Swain, and D.K. Sahoo, “Odia Text Summarization using Stemmer,” Int. Journal of Applied Information Systems (IJAIS), Vol. 1, No. 3, 2012. D.P. Sethi, “Design of Lightweight Stemmer for Odia Derivational Suffixes,” International Journal of Advanced Research in Computer and Communication Engineering, Vol. 2, No. 12, 2013. Odia grammar book of class 9th in BSE, ODISHA. A. Mohammad, S. Oqeili, and A. A. Rawan, “Occurrences Algorithm for String Searching Based on Brute-force Algorithm,” Jordan Journal of Computer Science, Vol. 2, No. 1, pp 82-85, 2006. D.P. Sethi, “Morphological Analyzer for Sambalpuri Odia Dialect Inflected Verbal Forms,” International Journal of Advanced Reseach in Computer Science and Sofiware Enginnering, Vol. 3, No. 10, 2013. M. Erritali, “Information Retrieval: Textual Indexing Using an Oriented Object Database,” Indonesian Journal of Electrical Engineering and Computer Science, Vol. 2, No. 1, pp.205-214, April 2016. I Gusti Ayu Triwayuni, I Ketut Gede Darma Putra, and I Putu Agus Eka Pratama, “Content Based Image Retrieval Using Lacunarity and Color Moments of Skin Diseases,” Indonesian Journal of Electrical Engineering and Computer Science, Vol. 9, No. 1, pp. 243-248, January 2018. P. Bajpai, P. Verma, Q. Abbas Syed, “Two Level Disambiguation Model for Query Translation,” International Journal of Electrical and Computer Engineering (IJECE), Vol. 8, No. 5, pp. 3923-3932, October 2018.

Design and implementation of an effective web-based hybrid stemmer for Odia … (Gouranga Charan Jena)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 20~26 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp20-26



Simplified down sampling factor based modified SVPWM technique for cascaded inverter fed induction motor drive Ravi Kumar Bhukya, P. Satish Kumar Department of Electrical Engineering, Osmania University, India

Article Info

ABSTRACT

Article history:

This paper presents a rivew, investigation and performance analysis of novel down samples factor based modified space vector PWM is called clamping SVPWM technique for cascaded Multilevel Invereter fed to Induction motor drive. In this paper the reference sine wave generated as in case of conventional off set injected SVPWM technique is modified by down sampling factor the reference wave by order of 10. The performance analyses of this modulation strategies are analyzed by apply for five level, seven level, nine level and eleven level inverter. The performance analysis of cascaded inverter interms of line voltage, stator current, speed, torque and total harmonic distortion. The results are depicting that PD PWM is more effective among the four proposed PWM technique. It is observed that the CSV Pulse width modulation ensures excellent, close to optimized pulse distribution results compared to SPWM technique and also 11-level inverter beter performance in case of low THD and better foundemental output voltages comapared to 5, 7, 9-level inverter. The proposed technique has been simulated using MATLAB/SIMULINK software. This proposed technique can be applied to N-level multilevel Inverter also.

Received Apr 22, 2019 Revised Nov 20, 2019 Accepted Jan 11, 2020 Keywords: CSVPWM Down sampling factor MSVPWM N-level cascaded inverter SPWM

This is an open access article under the CC BY-SA license.

Corresponding Author: Ravi Kumar Bhukya, Department of Electrical Engineering, University College of Engineering Osmania University, Hyderabad, Telangana, 500007, India. Email: rkpurnanaik2014@gmail.com

INTRODUCTION Multi-level diode clamped voltage fed inverters are recently becoming very popular for multimegawatt power applications. The main advantage of such an inverter topology is voltage division, i.e., the output voltage is produced through small steps of voltage, and therefore the individual switches are submitted only to these small voltages steps [1, 2]. The other advantages are low harmonic distortion at output, low dv/dt and extended range of under modulation. But it has the disadvantages like the increased number of switching devices and the complex control algorithm. Another important topology, named Cascade H-Bridge (CHB), has fewer components to achieve the same number of output voltage levels [3, 4]. In addition, several modulation and control strategies have been developed or adopted for multilevel inverters including the following: multilevel sinusoidal pulsewidth modulation (PWM), multilevel selective harmonic elimination, and space-vector modulation (SVM). The results of a patent search show that multilevel inverter circuits have been around for more than 25 years. An early traceable patent appeared in 1975, in which the cascade inverter was first defined with a format that connects separately dc-sourced fullbridge cells in series to synthesize a staircase ac output voltage [5].

Journal homepage: http://ijaas.iaescore.com

Int J Adv Appl Sci



ISSN: 2252-8814

In this paper presents proposed down sampling based clamping SVPWM control strategy of threephase five level, seven level, nine level and eleven level inverters are compared for THD. The paper mainly deals with the computation and the comparison of the motor harmonic losses of proposed CSV PWM solutions and with the selection of the solutions providing the best results. Finally, the drive harmonic losses will be compared for each levels. Special attention is dedicated to the latest and more relevant industrial applications of these inverter. Finally, the possibilities for future development are addressed. 2.

GENERALIZED DOWN SAMPLING FACTOR BASED CLAMPING SVPWM FOR CASCADED MULTILEVEL INVERTER The down sampling-based clamping SVPWM technique proposed by Lipo is based on SVM and the modification improved the vector sequences of the switching space. An offset voltage is required in the three phases’ references of the inverter, calculated by (1) to center the active vectors within the switching period. Different pulse width modulation strategies are used in multilevel medium and high-power conversion applications [6]. They can generally be classified into three categories such as Multistep, staircase frequency switching strategies, which synthesise the AC voltage by adding rectangular waveforms by means of the multilevel concept, and which often use pre calculated switching angles. Space vector PWM strategies, which have been extended from two level SVPWM technique and have been applied to three phase multilevel inverter. Carrier based PWM strategies the vertically shifted carrier scheme (LSCPWM) can be easily realizable on any digital controller. This scheme comes with three different techniques such as PD, POD and APOD. And the horzatically shifted carrier scheme as Phase Shifted Carrier PWM (PSCPWM) is the common PWM for cascaded MLI [7]. The main parameters of the modulation process are shown in Figure 1. In conventional SVPWM for multilevel inverters to find the switching time duration, for different inverter vectors, the mapping of the outer sectors to an inner sub hexagon sector is to be done. The switching inverter vectors corresponding to the concrete sectors are switched and the time periods premeditated from the mapped inner sectors. Implementing such a scheme in multilevel inverters will be very difficult, because higher number of sectors and inverter vectors are present. And in this method the computation time is increased for real time application. In carrier based PWM scheme aproper offset voltage is added to sinusoidal references before comparing with carrier waves, to attain the performance of a SVPWM [8] shown in Figure 2. 𝑉𝑎𝑠 = (𝑉𝑚 𝑆𝑖𝑛( 𝑤𝑡))

(1)

𝑉𝑏𝑠 = (𝑉𝑚 𝑆𝑖𝑛( 𝑤𝑡 − 120))

(2)

𝑉𝑐𝑠 = (𝑉𝑚 𝑆𝑖𝑛( 𝑤𝑡 + 120))

(3)

𝑉𝑜𝑓𝑓𝑠𝑒𝑡 =

(

)

(4)

After that we adding the three reference singals with Voffset voltage we generated reference singal shown in below equations Van, Vbn and Vcn. 𝑉𝑎𝑛 = (𝑉𝑎𝑠 + 𝑉𝑜𝑓𝑓𝑠𝑒𝑡)

(5)

𝑉𝑏𝑛 = (𝑉𝑏𝑠 + 𝑉𝑜𝑓𝑓𝑠𝑒𝑡)

(6)

𝑉𝑐𝑛 = (𝑉𝑐𝑠 + 𝑉𝑜𝑓𝑓𝑠𝑒𝑡)

(7)

In this paper, a simple technique to determine the offset voltage (To be added to the reference phase voltage for PWM generation for the entire modulation range) is presented, based onely on the sampled amplitudes of the reference phase voltages. The proposed modified reference PWM technique presents a simple way to determine the time instants at which the three reference phases cross the triangular carriers. To obtain the maximum possible peak amplitude of the fundamental phase voltage in linear modulation, the procedure for this is given in [9, 10]. After the modified SVPWM technique output, we can do the Down sampling of the order 10 each phase shown in Figure 2. Which is also sometimes called decimation, down sampling used for reduces the sampling rate and removes the samples from the signal. Whilst maintaining its length with respect to time. Some mathematically analysis to find out the down sampling factor shown in bellow, we can used descriptive time is 2e-6, sampling frequency is one by descriptive time (10 6/2), number of samples per phase is the ratio of sampling frequency by fundamental frequency such as 10k samples, Simplified down sampling factor based modified SVPWM technique for cascaded … (Ravi Kumar Bhukya)



ISSN: 2252-8814

in order reduced the down sampling by order of 10 such as 1k samples. By using the down sampling based modified space vector PWM technique such as PD, POD, APOD and PS. We can observe the reduced the THD in output line to line voltages. From (1), (2), (3), (4), (5), (6) and (7) we get generate reference wave compared to triangular carrier shown in Figure 3 to Figure 6. Is shown in the waveform results generated by adding the offset voltage described in with the reference sinusoidal waveform.

Figure 1. PDSPWM

Figure 2. Modified SVPDPWM

Figure 3. CSV PDPWM

Figure 4. CSV PODPWM

Figure 5. CSV APODPWM 3.

Figure 6. CSV PSPWM

SIMULATION RESULTS AND DISCUSSION Simulations have been carried out in MATLAB/Simulink environment for the 5-level, 7-level, 9-level and 11-level CMLI by implementing CSVPDPWM, CSVPODPWM, CSVAPODPWM and CSVPSPWM techniques. A 3-phase induction motor is considered as load for this scheme. Simulation results are analyzed by computing %THD and plotting harmonic spectra of different PWM techniques. The circuit arrangement for N-level CMLI is shown in Figure 7. This circuit can be used to implement all other PWM techniques related to SPWM, THIPWM and modified SVPWM. The necessary simulation parameters for CMLI are as follows: the total DC-link voltage for a phase- 400V, reference wave frequency-50Hz, and carrier frequency- 10KHz. The harmonic analysis of 5-level CMLI for CSVPDPWM, CSVPODPWM, CSVAPODPWM and CSVPSPWM technique with triangular carrier wave is shown in Figure 8 to Figure 11. The magnitude of fundamental component in the CSVPSPWM with triangular carrier technique produces more value of 430.4V. The CSVPDPWM with triangular carrier gives better total harmonic distortion of 17.10%. In the five-level single phase CMLI contain two H-bridges with series connections with the output phase voltage is 5-level and line voltage is 9-level of the inverter and we can obverse CSVPODPWM and CSVAPODPWM contain apporximetely value of THD and fundmental output voltages. Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 20 – 26

Int J Adv Appl Sci

ISSN: 2252-8814



Figure 7. Three phase N-level CMI inverter

Figure 8. Five level THD for CSV PDPWM

Figure 9. Five level THD for CSV PODPWM

Figure 10. Five level THD for CSV APODPWM

Figure 11. Five level THD for CSV PSPWM

The harmonic analysis of 7-level CMLI for CSVPDPWM, CSVPODPWM, CSVAPODPWM and CSVPSPWM technique with triangular carrier wave is shown in Figure 12 to Figure 15. The magnitude of fundamental component in the CSVAPODPWM with triangular carrier technique produces more value of 651.2V. The CSVPDPWM with triangular carrier gives better total harmonic distortion of 11.65%. In the 7-level single phase CMLI contain three H-bridges with series connections with the output phase voltage is 7-level and line voltage is 13-level of the inverter and we can obverse CSVPODPWM and CSVAPODPWM contain apporximetely value of THD and fundmental output voltages.

Figure 12. Seven level THD for CSV PDPWM

Figure 13. Seven level THD for CSV PODPWM

Simplified down sampling factor based modified SVPWM technique for cascaded … (Ravi Kumar Bhukya)



Figure 14. Seven level THD for CSV APODPWM

ISSN: 2252-8814

Figure 15. Seven level THD for CSV PSPWM

The harmonic analysis of 9-level CMLI for CSVPDPWM, CSVPODPWM, CSVAPODPWM and CSVPSPWM technique with triangular carrier wave is shown in Figure 16 to Figure 19. The magnitude of fundamental component in the CSVPDPWM with triangular carrier technique produces more value of 826.5V. The CSVPDPWM with triangular carrier gives better total harmonic distortion of 8.90%. In the 9level single phase CMLI contain four H-bridges with series connections with the output phase voltage is 9level and line voltage is 17-level of the inverter and we can obverse CSVPODPWM and CSVAPODPWM contain apporximetely value of THD and fundmental output voltages. The harmonic analysis of 11-level CMLI for CSVPDPWM, CSVPODPWM, CSVAPODPWM and CSVPSPWM technique with triangular carrier wave is shown in Figure 20 to Figure 23. The magnitude of fundamental component in the CSVPODPWM and CSVAPODPWM with triangular carrier technique produces more value of 1017 V. The CSVPDPWM with triangular carrier gives better total harmonic distortion of 6.68%. In the 11-level single phase CMLI contain four H-bridges with series connections with the output phase voltage is 11-level and line voltage is 21-level of the inverter and we can obverse CSVPODPWM and CSVAPODPWM contain apporximetely value of THD and fundmental output voltages. This proposed modified down sampling factor-based clamping SVPWM signal generation does not involve region identification, sector identification or look up tables for switching vector determination required in the conventional multilevel SVPWM technique. This scheme is computationally efficient when compared to conventional multilevel SVPWM scheme. We can observe that nuber of level is increased the total harmonic distortion is reduced and fundementel output voltage is increases shown in Table 1. The comparison to other conventional SVPWM technique and all other SPWM technique.

Figure 16. Nine level THD for CSV PDPWM

Figure 17. Nine level THD for CSV PODPWM

Figure 18. Nine level THD for CSV APODPWM

Figure 19. Nine level THD for CSV PSPWM

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 20 – 26

Int J Adv Appl Sci

ISSN: 2252-8814



Figure 20. Eleven level THD for CSV PDPWM

Figure 21. Eleven level THD for CSV PODPWM

Figure 22. Eleven level THD for CSV APODPWM

Figure 23. Eleven level THD for CSV PSPWM

Table 1. Comparison of THD for different level of the cascaded inverter fed induction motor Output voltage levels Five-level

Seven-level

Nine-level

Eleven-level

Techniques CSV-PDPWM CSV-PODPWM CSV-APODPWM CSV-PSPWM CSV-PDPWM CSV-PODPWM CSV-APODPWM CSV-PSPWM CSV-PDPWM CSV-PODPWM CSV-APODPWM CSV-PSPWM CSV-PDPWM CSV-PODPWM CSV-APODPWM CSV-PSPWM

THD (%) @ Fundamental Output Voltage 17.10 (325.9) 21.61 (390.3) 21.54 (390.3) 23.48 (430.4) 11.65 (518.5) 16.26 (651.1) 16.58 (651.2) 19.89 (632.2) 8.90 (826.5) 9.35 (825.4) 9.03 (826.5) 19.17 (785.8) 6.68 (1009) 8.65 (1017) 8.79 (1017) 13.27 (973.5)

CONCLUSION In this paper dealy with a novel down sampling factor based modified SVPWM technique so called Clamping Space vector Pulse width modulation (CSVPWM) technique. The reference sine wave generated as in case of conventional off set injected SVPWM technique is modified by down sampling the reference wave by order of 10. The comparison of THD of the proposed control strategies for 5, 7, 9 and 11-level inverter. When compared, it is obvious that CSV-PDPWM is the most efficient control strategy with low THD and increases the fundamental output voltages. The THD analysis, line voltages, stator currents and speed and torque of the machine are calibrated and compared confirming the good-quality waveforms. APPENDIX Table 1. Specification of induction motor Parameters Input voltage Inverter voltage Rotor speed Fundamentalfrequency Switching frequency Reference speed Frequency modulation Amplitude modulation Sampling factor order

Specifications 400VRMS(PhasePhase) 100(Volts) 1440(RPM) 50(Hz) 10K (Hz) 1500(RPM) 200 0.866 10

Simplified down sampling factor based modified SVPWM technique for cascaded … (Ravi Kumar Bhukya)



ISSN: 2252-8814

ACKNOWLEDGMENT We thank the University Grants Commission (UGC), Govt. of India, New Delhi for providing Major Research Project to carry out the Research work on Multi-Level Inverters. I also thank UGC for awarding me with RGNF FELLOWSHIP to carry out my research work, Department of Electrical Engineering, University Colleges of Engineering, Osmania University (Ph. D). REFERENCES [1]

Akira Nabae, Isao Takahashi, and Hirofumi Akagi, “A New Neutral-Point-Clamped Pwm Inverter,” IEEE Trans on Industry Applications, vol. ia-17, no. 5, pp 518-523, 1981. [2] Irfan Ahmed and Vijay B. Borghate, “Simplified space vector modulation technique for seven-level cascaded Hbridge inverter,” IET Power Electron, vol. 7, no. 3, pp. 604–613, 2014. [3] José Rodriguez, et al, “Multilevel Inverters: A Survey of Topologies, Controls, and Applications,” IEEE Transactions on Industrial Electronics, vol. 49, no. 4, pp 724-738, Aug 2002. [4] Ravi Kumar Bhukya, P. Satish Kumar, and E. Sreenu, “Analysis of level shifted modulation strategies applied to cascaded H-bridge multilevel inverter fed induction motor drive,” Sixth International Conference on Advances in Computing, Control and Networking–ACCN 2017, pp. 80-84, 2017. [5] J.-S. Lai and F.Z. Peng, “Multilevel converters - a new breed of power converters,” IEEE Trans-actions on Industry Applications, vol. 32, no. 3, pp. 509-517, 1985. [6] A. Gupta and A. Khambadkone, “A space vector PWM scheme for multilevel inverters based on two-level space vector PWM,” IEEE Trans. Ind. Electron., vol. 53, no. 5, pp. 1631–1639, Oct. 2006. [7] F. Z. Peng and J. S. Lai, “Multilevel cascade voltage-source inverter with separate DC sources,” U.S. Patent 5,642,275, June 1997. [8] D. Asadei, G. Serra, and K. Tani, “Implementation of a Direct Control Algorithm for Induction Motors Based on Discrete Space Vector Modulation,” IEEE Transactions on Power Electronics, vol. 15, no. 4, pp. 769-777, 2000. [9] M. Satyanarayana and P. Satish Kumar, “Analysis and Design of Solar Photo Voltaic Grid Connected Inverter,” Indonesian Journal of Electrical Engineering and Informatics (IJEEI), vol. 3, no. 4, pp 199-208, December 2015. [10] Ravi Kumar Bhukya and P. Satish Kumar, “Performance Analysis of Modified SVPWM Strategies for Three Phase Cascaded Multi-level Inverter fed Induction Motor Drive,” International Journal Of Power Electronics and Drive Systems (IJPEDS), vol. 8, no. 2, pp. 835-843, May 2017.

BIOGRAPHIES OF AUTHORS Mr. Ravi Kumar Bhukya was born in Mahabubabad (Warangal), Telangana, India. He obtained B.Tech. in Electrical and Electronics Engineering from JNTU University, Hyderabad in 2010 and M.Tech. in Power and induristal drives Engineering in 2013 from Jawaharlal Nehru Technological University, Hyderabad. His research interests include Power Electronics, Drives, Power converters and Multi level inverter. Presently he is pursuing Ph. D. in Osmania University, Hyderabad, INDIA.

Dr. Satish Kumar Peddapelli is an Associate Professor in the Department of Electrical Engineering, University College of Engineering, Osmania University, Hyderabad. He has completed his B.Tech in EEE from JNTU, obtained his M.Tech in Power Electronics from JNTUH and his Doctorate in the area of Multilevel Inverters in the year 2011 from JNTUH. His areas of interests are Power Electronics, Drives, Power Converters, Multi Level Inverters, Special Machines and Hybrid Power Systems. He is the Principal Investigator for three Major Research Projects funded by UGC, SERB, worth around 9 Lakhs, 21 Lakhs and Indo-Sri Lanka joint research project worth of around 25 lakhs from the Department of Science and Technology, New Delhi. He has more than 60 publications in International Journals and has attended and presented papers in 28 International Conferences. He authored one text book. He received the “Best Teacher award” from the state Government of Telangana on 5 th September, 2014.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 20 – 26

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 27~33 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp27-33



An efficient quantum multiverse optimization algorithm for solving optimization problems Samira Sarvari, Nor Fazlida Mohd. Sani, Zurina Mohd Hanapi, Mohd Taufik Abdullah Department of Computer Science, Faculty of Computer Science and Information Technology, Universiti Putra Malaysia, Malaysia

Article Info

ABSTRACT

Article history:

Due to the recent trend of technologies to use the network-based systems, detecting them from threats become a crucial issue. Detecting unknown or modified attacks is one of the recent challenges in the field of intrusion detection system (IDS). In this research, a new algorithm called quantum multiverse optimization (QMVO) is investigated and combined with an artificial neural network (ANN) to develop advanced detection approaches for an IDS. QMVO algorithm depends on adopting a quantum representation of the quantum interference and operators in the multiverse optimization to obtain the optimal solution. The QMVO algorithm determining the neural network weights based on the kernel function, which can improve the accuracy and then optimize the training part of the artificial neural network. It is demonstrated 99.98% accuracy with experimental results that the proposed QMVO is significantly improved optimization compared with multiverse optimizer (MVO) algorithms.

Received Jun 29, 2019 Revised Nov 2, 2019 Accepted Dec 1, 2019 Keywords: Intrusion Detection System Multiverse Optimization Quantum Computing Quantum Multiverse Optimization

This is an open access article under the CC BY-SA license.

Corresponding Author: Samira Sarvari, Department of Computer Science, Faculty of Computer Science and Information Technology, Universiti Putra Malaysia, Serdang, 43400 Seri Kembangan, Selangor, Malaysia. Email: samirasarvari82@yahoo.com

INTRODUCTION The increase in the number of local networks has led to the continuous development of Internet data and the availability of massive amounts of network data has promoted the development of information technology, which requires careful attention. As a result, this evolution, in turn, has increased the system's vulnerability to various threats [1]. Any intrusion can have catastrophic consequences. For example, personal data may be destroyed, corrupted or illegally accessed as a result of breaches of confidentiality. In addition, infringements of integrity can lead to alteration of personal data. Computer network security has become a promising tool for secure channels. One of the promising tools for detecting attacks is the intrusion detection system (IDS). Cybersecurity infrastructures use IDS as an essential component and protect systems and infrastructures against various threats. An intrusion detection system consists of data collection, data clearing and pre-processing, intrusion detection, reporting and reasonable action, which is an essential part of these attack detection processes [2]. High classification accuracy and a low false alarm rate are the two main characteristics of well-developed IDS, so it is extremely important to develop mechanisms for intrusion detection in view of the conviction that suspicious activities can be detected by taking measures to prevent further breeding of computer networks or systems [3]. Data classification has been studied extensively in many computer fields and up to now, the development of classification has achieved great achievements and many types of classified technology Journal homepage: http://ijaas.iaescore.com



ISSN: 2252-8814

and theory will continue to emerge. In the face of a lot of noisy, cluttered, nonlinear data, artificial neural network (ANN) not only helps to make high - quality modeling and complete training in the process of using large amounts of data, but also has a test mode set to evaluate the performance of ANN. ANNs are a form of machine learning algorithm inspired by the behavior of biological neurons in the brain and central nervous system, and use mathematical models to describe the architecture of biological neural networks to solve information processing problems. The ANN model compromises three layers: the input layer, the hidden layers and the output layer. The weights connecting the input layer to the hidden layer, as well as the bias values of the hidden layer, are randomly generated before the learning process. Only the weights connecting the hidden layer to the output layer are trained by the fast-linear regression. An example of a simple ANN with a single hidden layer is shown in Figure 1, where "I" is the input of the neural network and "W" is the weight given to each input.

Figure 1. Simple ANN with a single hidden layer Recently, researchers from all over the world have been improving the ANN according to different forecasting tasks and have obtained some satisfying results [4]. Nevertheless, the gradient-based learning algorithms are widely used to train traditional ANNs, which may result in some drawbacks such as the slow convergence speed, the local minimum, and the overfitting phenomenon. In order to solve the aforementioned problems, we focus our research on an improved machine learning algorithm based on neural networks with random weights and kernels (KNNRW). Recently, neural networks with random weights and kernels (KNNRW) [5] has been proposed by replacing the hidden nodes mapping with the kernel mapping. It does not need to determine the number of hidden nodes of KNNRW. In order to solve the above problems, we focus our research on an improved neural network with random weights and kernels (KNNRW) and proposed quantum multiverse optimization (QMVO) algorithm. In this research, the neural network algorithm is first improved by determining the neural network weights based on the kernel function, and then optimize the training part of the artificial neural network to develop advanced detection approach for IDS. 2.

BASICS AND BACKGROUND In the last years, bio-inspired computing has witnessed advances, popularity, and interest in different areas of sciences and engineering. However, there are still some problems remaining unsolved. Some of these are classifications methodology, parameters tuning, the gap between theoretical and practical parts, largescale real-world applications, and finally the selection of the appropriate algorithm for specific problem [6]. The motivation of the hybrid algorithms to solve the optimization problem is based on No Free Lunch (NFL) theorem [7]. According to NFL theorem, no algorithm is able to solve all the optimization problems. Therefore, this topic of research is open until now. Thus, researchers make many efforts to improve the current optimization algorithms to solve different complex problems. Some of them used the benefits of quantum computing (QC) to solve these problems such as the speed, efficiency, and performance of evolutionary algorithms. QC is a new emerging mechanism in computer science and engineering and other disciplines. QC is a branch of mathematics that uses the specificities of quantum mechanics for data transformation and information processing, which stored in a two-state quantum bits or qubit. It has gained the interest of researchers in the last years in the fields such as quantum algorithms and quantum computers. Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 27 – 33

Int J Adv Appl Sci

ISSN: 2252-8814



QC depends on some principles of quantum mechanics in which the smallest information unit is called the quantum bit or qubit. QC uses 0 and 1 for representing the two basic states. The main difference between bit and qubit is that the qubit can be in a state between 0 and one not only in a state of 0 or 1 as in the classical bit. In QC, the processing of the enormous number of quantum states is implemented in a parallel way simultaneously [8]. Therefore, many types of research have studied the theoretical and practical studies to merge quantum computing and evolutionary computation [9]. Some of these are quantum genetic algorithms [10], quantum inspired scatter search [9, 11], quantum differential algorithm [12], mining large databases [13], 0–1 optimization problem [14], knapsack problem [15], traveling salesman problem [16], engineering inverse problem [17] and other areas of applications as in Gottfried and Yan [18]. Currently, quantum inspired algorithms are used to solve many combinatorial optimization problems as proposed [19]. There are many hybridized quantum evolutionary algorithms proposed in the literature such as quantum inspired evolutionary algorithms [10], quantum inspired immune algorithm [20], quantum PSO [21]. Sun et al. [21], quantum inspired PSO employed a probability searching technique, and the search space is transferred from classical space to quantum, where the particles’ movement is similar to the ones with the quantum mechanics [22]. Quantum inspired evolutionary algorithms were introduced to solve the traveling salesman problem [23], where the crossover operation was per-formed based on the interference concept. Multiverse optimizer (MVO) is one of the bio-inspired algorithms [24]. The main inspiration of MVO is taken from multiverse theory in physics. It is based on three main concepts in cosmology. These concepts are a wormhole, white hole, and black hole. Like other evolutionary algorithms, MVO starts the optimization process by creating a population of solutions. In fact, this algorithm mimics the interaction between multiple universes through the wormhole, black hole, and white hole. The core idea of MVO came from the fact that larger universes tend to send objectives to smaller universes to reach a stable position. MVO has been used to solve many optimization problems. Nineteen unimodal/multimodal benchmark functions have been adopted to evaluate the performance of MVO [24]. Fariset et al. [25] used MVO to select the optimal feature subset. Moreover, they used it to optimize SVM parameters. Han and Kim [10], employed the MVO for training the multilayer perceptions neural network. Their approach was evaluated and benchmarked using nine different biomedical datasets selected from the UCI machine learning repository. To assess the performance of this algorithm, the obtained results are compared with ﬁve recent evolutionary meta-heuristic algorithms: particles warm optimization (PSO), genetic algorithm (GA), differential evolution (DE) algorithm, and cuckoo search (CS). These studies have been revealed that MVO algorithm is more efﬁcient than the other algorithms, and it can obtain better results. However, MVO algorithm like most of the optimization algorithms suffers from low exploitation and convergence rate. Zouache and Moussaoui [15] proposed a hybrid algorithm based on using particles warm optimization (PSO) and multiverse optimization (MVO). They combine the exploitation capability of PSO and the exploration capabilities of MVO. The experimental results validate its effectiveness compared to standard PSO and MVO. There is always room for improvement and adaptation of an algorithm to solve a particular set of problems. 3.

MATHEMATICAL MODELING Neural networks with weights and kernels (KNNRW) have been proposed by introducing the kernel function mapping of SVM as the hidden node mapping of NNRW. The optimization problem of NNRW can be written as (1) min 𝐿

‖𝑤‖ +

∑

(1)

‖𝜉 ‖

ℎ(𝑥 ) . 𝑤 = 𝑡 − 𝜉 , 𝑖 = 1, … , 𝑁 Where 𝜉 is the training error related to the ith training sample 𝑥 , C is the regularization coefficient, and ℎ(𝑥 ) denotes the ith row of H. The corresponding dual optimization problem of (1) can be formulated as (2) 𝐿

‖𝑤‖ +

∑

‖𝜉 ‖ − ∑

𝛼 (ℎ(𝑥 ) . 𝑤 − 𝑡 + 𝜉 )

(2)

Where 𝛼 is the langrage multiplier with respect to the ith training sample 𝑥 . The corresponding Karush-Kuhn-Tucker (KKT) conditions are as follows =0 ⟶𝑤= ∑

𝛼 ℎ(𝑥) ⟶ 𝑤 = 𝐻 𝛼

(3)

An efficient quantum multiverse optimization algorithm for solving optimization problems (Samira Sarvari)



ISSN: 2252-8814 = 0 ⟶ 𝛼 = 𝐶𝜉 , 𝑖 = 1, … , 𝑁

(4)

= 0 ⟶ ℎ(𝑥 ) . 𝑤 − 𝑡 + 𝜉 = 0, 𝑖 = 1, … , 𝑁

(5)

Substituting (3) and (4) into (5), the following equation can be obtained + 𝐻𝐻

(6)

𝛼=𝑇

Where I is an identity matrix. Considering (3) and (6), the weight w can be calculated as 𝑤=𝐻

+ 𝐻𝐻

(7)

𝑇

Thus, the output function of NNRW can be written as 𝑓(𝑥) = ℎ(𝑥)𝐻

+ 𝐻𝐻

(8)

𝑇

It can be seen from (8) that the specific form of h(x) is not important as long as the dot product of 𝐻𝐻 (or ℎ(𝑥)𝐻 ) is known. As a result, if the hidden node mapping h(x) is unknown, we can define the kernel matrix of KNNRW as follows Ω Ω

= 𝐻𝐻 ∶ = ℎ(𝑥 ) . ℎ 𝑥

(9)

= 𝐾 𝑥 ,𝑥

Consequently, the output function can be rewritten accordingly as 𝐾(𝑥, 𝑥 ⎡ . ⎢ . 𝑓(𝑥) = ⎢ . ⎢ ⎣𝐾(𝑥, 𝑥

)

⎤ ⎥ ⎥ ⎥ )⎦

+Ω

𝑇

(10)

In order to further improvement of accuracy, we adjust and update the weights in ANN using the recently proposed Quantum multiverse optimization (QMVO) algorithm. In QMVO algorithm, each universe has a state depicted by wave function 𝜓(𝑦, 𝑡) with probability density function of universe position |𝜓(𝑦, 𝑡)| . In quantum, the dynamic behavior of universe differs from the standard version of MVO. The updating position is mathematically formulated as follows (11) Where 𝛿(𝑦) is the Dirac delta function and k is a positive value. The universe wave function in delta potential is deﬁned as follows:

(12) Where m denotes the universe’s mass and h denotes the reduced plank constant. The particles (universes) are deﬁned by the following (13)

(13) Where u and l are randomly initialized in interval [0, 1], 𝛽 denotes the contraction–expansion coefﬁcient. It is linearly decreased over iteration. The mathematical formula is deﬁned as follows Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 27 – 33

Int J Adv Appl Sci

ISSN: 2252-8814



(14) Where 𝑀𝑎𝑥 defined as the maximum number of iterations and iter denotes the current iteration number. Mbest is defined as the population’ mean best or global point. It is mathematically defined as follows

(15) Where M is the number of universes (population size), g is the best universe 𝛼 index among all universes in the search space, dim denotes the universe dimension and MaxDim is the maximum number of dimensions. Through this, the local attractor in order to guarantee the convergence speed of the QMVO is deﬁned as follows

(16) Where 𝑟 and 𝑟 are the random numbers in range [0, 1], 𝑥 , is the ith universe index in dimth dimensions of the hyperspace and 𝑥 , denotes the gth best universe position index of dimth dimensions. The whole algorithm steps of QMVO are presented at algorithm. Quantum Multi Verse Optimization Algorithm (QMVO) 1: Randomly set the initial values of the universes’s positions (population size) M, dimensions MaxDim, lowerlbp and upper ubp boundaries, the maximum number of iterations 𝑀𝑎𝑥 and the best universe 𝛼 . 2: SU=Sorted universes 3: Normalize the fitness value (the inflation rate) of the universes NI. 4: Set iter: = 1. {Counter initialization}. 5: repeat 6: Calculate the fitness value (the inflation rate) of the universes. 7: Check if the new universes positions go out of the search space boundaries and bring it back. 8: for (i =1: i <= M) do 9: Update Mbest and ß using (14) and (15) 10: Set black hole index to i. 11: for (j =1: i < = size of the universe position) do= 12: 𝑅 = random ([0, 1]) 13: if 𝑅 <= NI ( 𝑦 ) then 14: White hole index = Roulette Wheel Selection(-NI); 15: U (black hole index, j) =SU (white hole index, j) 16: end if 17: 𝑅 = random ([0, 1]); 18: if 𝑅 < Wormhole probability existence then 19: l = random ([0, 1]); 20: u = random ([0, 1]); 21: if l < 0.5 then 22: 𝑦 (t +1) = 𝛼 – ß. |𝑀 𝑏𝑒𝑠𝑡 − 𝑦 (𝑡)|. 𝑙𝑛 1 𝑢 ; 23: else 24: 𝑦 (t +1) = 𝛼 – ß. |𝑀 𝑏𝑒𝑠𝑡 − 𝑦 (𝑡)|. 𝑙𝑛 1 𝑢 ; 25: end if 26: end if 27: end for 28: end for 29: Set iter = iter +1. {Iteration counter increasing} 30: until (iter < 𝑀𝑎𝑥 ) . {Termination criteria satisfied}. 31: Produce the best universe.

An efficient quantum multiverse optimization algorithm for solving optimization problems (Samira Sarvari)



ISSN: 2252-8814

The training process is an important phase for the optimization of the ANN. In the ANN training, each individual represents all of the weights and biases of the ANN structure. The objectives of this process are to search for the synaptic weights of the ANN and to reduce the MSE, which represents the cost function of the ANN and achieve the highest classification and prediction accuracy. The training of the neural network is determined by the QMVO algorithm based on the weights and the error rate obtained is based on the following pseudo-code. 1: initialize all universe as input parameter 2: for i=1 until count of universe 1: calculate 𝑢𝑛𝑖𝑣𝑒𝑟𝑠𝑒 MSE 3: if 𝑢𝑛𝑖𝑣𝑒𝑟𝑠𝑒 MSE are the minimum vale in all MSE universe 4: neural network weight s is 𝑢𝑛𝑖𝑣𝑒𝑟𝑠𝑒 5: best universe is 𝑢𝑛𝑖𝑣𝑒𝑟𝑠𝑒 6: endif 7: endfor 8: output bet universe is weight neural network Based on this pseudo-code, all worlds are considered as inputs of the system first, then the values of these worlds are considered as weights of the neural network. The MSE value or the error of each neural network is calculated according to the weights of each universe. Accordingly, the world with the lowest MSE rate is considered as the weights of the final neural network. In this way, the neural network is trained at each stage of the QMVO algorithm. 4.

RESULTS AND DISCUSSION The proposed model was implemented and evaluated in MATLAB. In the QMVO+ANN training using the NSL-KDD dataset with number of iterations=100. The IDS was evaluated on several factors. The main factors included in this research are detection rate (DR), false alarm rate (FAR), and accuracy (ACC). In the literature, most of research works in the ﬁeld of intrusion detection focused on the accuracy, the detection rate (DR), and false alarm rate (FAR). In this research, we have adopted the same metrics to evaluate the performance of our proposed approach. The experimental results in Figure 2 provides an evaluation of the performance of the ANN intrusion detection for the QMVO algorithm with 97.48 DR, 0.03 FAR, and 99.98 ACC. The results show the potential applicability of QMVO with ANN for developing practical IDSs. Further-more, the comparisons of the performance results of the proposed ANN+QMVO and another model ANN+MVO dataset are shown in Figure 3. The proposed ANN+QMVO clearly performs the best in terms of ACC and DR. The data correctly classified by the proposed ANN+QMVO are more than correctly classified by ANN+MVO.

Figure 2. Combination of ANN and QMVO

Figure 3. Comparison of ANN+MVO and ANN+QMVO

CONCLUSION There are various techniques of Artificial Neural Network, which can be applied to intrusion detection system. Each technique is suitable for some specific situation. QMVO is easy to implement, supervised learning artificial neural network. Number of the epochs required to train the network is high Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 27 – 33

Int J Adv Appl Sci

ISSN: 2252-8814



as compare to the other ANN techniques. This research presents a new global optimization algorithm called quantum multiverse optimization (QMVO) with ANN and quantum behavior for solving the optimization problems. Therefore, the proposed method offers the best strong exploration and precise exploitation capabilities. In this research, the proposed QMVO algorithm was only implemented for solving optimization problems. Thus, our future work will concentrate on implementing the QMVO in (i) solving more complex optimization problems with different properties and (ii) design conceptions for engineering, practical applications, and constrained problems. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25]

I. Ahmad, A. B. Abdullah, and A. S. Alghamdi, "Application of artificial neural network in detection of probing attacks," in Industrial Electronics & Applications, 2009. ISIEA 2009. IEEE Symposium, pp. 557-562, 2009. S. X. Wu and W. Banzhaf, "The use of computational intelligence in intrusion detection systems: A review," Applied soft computing, vol. 10, pp. 1-35, 2010. S. M. H. Bamakan, B. Amiri, M. Mirzabagheri, and Y. Shi, "A new intrusion detection approach using PSO based multiple criteria linear programming," Procedia Computer Science, vol. 55, pp. 231-237, 2015. Z. He, Q. Hu, Y. Zi, Z. Zhang, and X. Chen, "Hybrid intelligent forecasting model based on empirical mode decomposition, support vector regression and adaptive linear neural network," in International Conference on Natural Computation, pp. 324-327, 2005. G.-B. Huang, "An insight into extreme learning machines: random neurons, random features and kernels," Cognitive Computation, vol. 6, pp. 376-390, 2014. X.-S. Yang and M. Karamanoglu, "Swarm intelligence and bio-inspired computation: an overview," in Swarm Intelligence and Bio-Inspired Computation, ed: Elsevier, pp. 3-23, 2013. D. H. Wolpert and W. G. Macready, "No free lunch theorems for optimization," IEEE transactions on evolutionary computation, vol. 1, pp. 67-82, 1997. A. Montanaro, "Quantum algorithms: an overview," npj Quantum Information, vol. 2, p. 15023, 2016. A. Layeb and D. E. Saidouni, "A new quantum evolutionary local search algorithm for MAX 3-SAT problem," in International Workshop on Hybrid Artificial Intelligence Systems, pp. 172-179, 2008. K. H. Han and J. H. Kim, "Quantum-inspired evolutionary algorithms with a new termination criterion, h-epsilon gate, and two-phase scheme," IEEE transactions on evolutionary computation, vol. 8, pp. 156-169, 2004. A. Layeb, "Hybrid quantum scatter search algorithm for combinatorial optimization problems," Journal of Annals. Computer Science Series, vol. 8, pp. 227-244, 2010. A. Draa, S. Meshoul, H. Talbi, and M. Batouche, "A quantum-inspired differential evolution algorithm for solving the N-queens problem," Neural networks, vol. 1, 2011. M. Ykhlef, "A quantum swarm evolutionary algorithm for mining association rules in large databases," Journal of King Saud University-Computer and Information Sciences, vol. 23, pp. 1-6, 2011. A. Layeb, "A hybrid quantum inspired harmony search algorithm for 0–1 optimization problems," Journal of Computational and Applied Mathematics, vol. 253, pp. 14-25, 2013. D. Zouache and A. Moussaoui, "Quantum-Inspired Differential Evolution with Particle Swarm Optimization for Knapsack Problem," J. Inf. Sci. Eng., vol. 31, pp. 1757-1773, 2015. B. A. L. d. M. Herrera, L. d. S. Coelho, and M. T. A. Steiner, "Quantum inspired particle swarm combined with Lin-Kernighan-Helsgaun method to the traveling salesman problem," Pesquisa Operacional, vol. 35, pp. 465-488, 2015. O. U. Rehman, S. Yang, and S. U. Khan, "A modified quantum-based particle swarm optimization for engineering inverse problem," COMPEL-The International Journal For Computation And Mathematics In Electrical And Electronic Engineering, vol. 36, pp. 168-187, 2017. K. Gottfried and T.-M. Yan, Quantum mechanics: fundamentals: Springer Science & Business Media, 2013. A. Layeb and D.-E. Saidouni, "A new quantum evolutionary algorithm with sifting strategy for binary decision diagram ordering problem," International Journal of Cognitive Informatics and Natural Intelligence (IJCINI), vol. 4, pp. 47-61, 2010. L. Jiao, Y. Li, M. Gong, and X. Zhang, "Quantum-inspired immune clonal algorithm for global optimization," IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), vol. 38, pp. 1234-1253, 2008. J. Sun, B. Feng, and W. Xu, "Particle swarm optimization with particles having quantum behavior," in Evolutionary Computation, 2004. CEC2004. Congress, pp. 325-331, 2004. K.-L. Du and M. Swamy, "Search and optimization by metaheuristics," Birkhaüser, Jul 2016. A. Narayanan and M. Moore, "Quantum-inspired genetic algorithms," in Evolutionary Computation, 1996 Proceedings of IEEE International Conference, pp. 61-66, 1996. S. Mirjalili, S. M. Mirjalili, and A. Hatamlou, "Multi-verse optimizer: a nature-inspired algorithm for global optimization," Neural Computing and Applications, vol. 27, pp. 495-513, 2016. H. Faris, M. A. Hassonah, A.-Z. Ala’M, S. Mirjalili, and I. Aljarah, "A multi-verse optimizer approach for feature selection and optimizing SVM parameters based on a robust system architecture," Neural Computing and Applications, vol. 30, pp. 2355-2369, 2018.

An efficient quantum multiverse optimization algorithm for solving optimization problems (Samira Sarvari)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 34~42 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp34-42



Trilateration based localization method using mobile anchor in wireless sensor networks M. G. Kavitha1, K. Vinoth Kumar2,T. Jayasankar3 1Department

of CSE, University College of Engineering, India of ECE, SSM Institute of Engineering and Technology, India 3Department of ECE, University College of Engineering, India

2Department

Article Info

ABSTRACT

Article history:

Localization in wireless sensor networks (WSNs) is essential in many applications like target tracking, military applications and environmental monitoring. Anchors which are equipped with global positioning system (GPS) facility are useful for finding the location information of nodes. These anchor nodes may be static or dynamic in nature. In this paper, we propose mobile anchors assisted localization algorithm based on regular hexagons in two-dimensional WSNs. We draw a conclusion that the number of anchor nodes greatly affect the performance of localization in a WSN. An optimal number of anchor nodes significantly reduces the localization error of unknown nodes and also guarantees that unknown nodes can obtain high localization accuracy. Because of the mobility of anchor nodes high volume of sensing region is covered with less period of time and hence the coverage ratio of the proposed algorithm increases. Number of communications also decreases for the reason that the system contains loge (n) number of anchor nodes which leads to less energy consumption at nodes. Simulation results show that our LUMAT algorithm significantly outperforms the localization method containing single anchor node in the network. Movement trajectories of mobile anchors should be designed dynamically or partially according to the observable environment or deployment situations to make full use of realtime information during localization. This is the future research issue in the area of mobile anchor assisted localization algorithm.

Received Apr 23, 2019 Revised Oct 25, 2019 Accepted Dec 20, 2019 Keywords: Localization LUMAT Mobile anchor nodes Wireless sensor networks

This is an open access article under the CC BY-SA license.

Corresponding Author: K. Vinoth Kumar, Department of ECE, SSM Institute of Engineering and Technology, Dindigul, Tamil Nadu, India-624 002. Email: vinodkumaran87@gmail.com

INTRODUCTION A sensor network comprises of a large number of sensor nodes that are densely deployed in a field. Each sensor performs sensing task for detecting specific events. The sink node is responsible for collecting sensed data reported from all the sensors, and finally transmits the data to a task manager. If the sensors cannot directly communicate with the sink, some intermediate sensors performs the operation of forwarding the data to sink [1]. Wireless Sensor Networks (WSNs) have emerged as one of the key enablers in recent years for a variety of applications such as environment monitoring, vehicle tracking and mapping, and emergency response. One important problem in such applications is finding the position of a node. To solve the localization problem, it is natural to consider placing sensors manually or equipping each sensor with a GPS receiver. Constraints such as cost and power consumption make these two methods inefficient in the network, especially Journal homepage: http://ijaas.iaescore.com

Int J Adv Appl Sci

ISSN: 2252-8814



for large-scale WSNs. Hence a variety of approaches have been devised for sensor network localization [2]. Location discovery is emerging as one of the more important tasks as accurate location information could greatly improve the performance of tasks such as routing, energy conservation, data aggregation and maintaining network security [3, 4]. Localization in wireless sensor networks is performed in following steps. First, distance estimation: this phase involves measurement techniques to estimate the relative distance between nodes. Position computation: it consists of algorithms to calculate the coordinates of the unknown node with respect to the location of known anchor nodes or other neighboring nodes. Localization algorithms require techniques for location estimating depending on the beacon nodes’ location. These are called multi-lateration (ML) techniques. Iterative ML: Some nodes may not be in the direct range of three beacons. Once a node estimates its location, it sends out a beacon, which enables some other nodes to now receive atleast three beacons. Iteratively, all nodes in the network can estimate their location but location estimation may not be accurate as errors may propagate. Collaborative MLis when two or more nodes cannot receive atleast three beacons each, they collaborate with each other. Figure 1 shows nodes A and B have three neighbors each. Of the six participating nodes, four are beacons, whose positions are known. Proximity technique is used when there is no range information available. It reveals whether or not a node is in range or near to a reference point. Localization algorithms using this technique determine if a node is in proximity to a reference point by enabling the reference to transmit periodic beacon signals and whether the node is able to receive at least certain value of the beacon signals set as threshold. In a period, t if it receives n beacons greater than the set threshold then it is in proximity to that reference point [5, 6]. Localization algorithms: it determines how the information concerning distances and positions, is manipulated in order to allow most or all nodes of WSN to estimate their position. Optimally the localization algorithms may involve algorithms to reduce the errors.

Figure 1. Network architecture

In this paper, we propose a mobile anchor assisted localization algorithm based on Trilateration method (LUMAT) with the objectives of maximizing localization ratio, energy efficiency and localization accuracy. LUMAT uses loge (n) number of mobile anchor nodes as the reference nodes, which move in the sensing field and broadcast their current position periodically. Sensor nodes receive the position information of the mobile anchor nodes and localize themselves by using Trilateration algorithm. The results of simulations and measurements show that LUMAT is a practical method that can be used in real-world system, and is also a method of principle simple, less computing and communication, low cost, and high accuracy. The rest of the paper is organized as follows: the next section surveys related works on previous localization research, especially about methods based on mobile localization with more details in order to clarify our work. In section III, we describe the LUMAT method. Section IV reports our simulation and experimental results. Finally, we present our conclusion in section V.

RELATED WORK A general survey about localization for wireless sensor networks is found in which is a broad research area in the past several years. A brief survey about various range-free approaches and localization methods which involve mobile reference nodes are provided here. Energy consumption phenomenon has always been noticed in sink-based wireless sensor networks. This paper explores an energy efficient routing protocol with a mobile sink based on the shortest path data transmission mode [7]. According to the position of the sink node and the common nodes' ID in the network, we calculate the coordinate value of each node in the network. Trilateration based localization method using mobile anchor in wireless sensor networks (M. G. Kavitha)



ISSN: 2252-8814

By comparing the coordinate values to choose the shortest path to forward the data. Simulation results show that this method can prolong the lifetime of the network, improve the energy utilization ratio and the energy consumption more balanced. Clustering methodology play a significant role in improving the lifetime of sensor network. There has been various hierarchical clustering that has been developed in recent time by enhancing the protocol [8]. The drawback of these protocols is energy of cluster head degrades very fast due to long distance transmission and packet failure likelihood is not considered for inter and intra cluster transmission. To address the energy efficiency issue of existing approach this work proposed packet failure likelihood estimation model and hop selection optimization model for inter cluster transmission. Positional accuracy is very important indicator for assessing the location of performance. More localization is high precision location of the performance is better [9]. A conclusion might elaborate on the importance of the work or suggest applications and extensions. In addition, the accuracy of the location of the amorphous algorithm is superior to that of other algorithms and there is not a large increase of energy consumption, which is why it is suitable for the location of network nodes large scale. Enhancing the lifetime of the sensor network and at the same time maintaining proper security is an important aspect in wireless sensor networks [10]. In this paper we use the concept of clustered wireless sensor networks. Clustering is a key concept for enhancing the sensor network’s lifetime. Cluster wireless sensor networks have mainly two benefits than non clustered WSNs. They are: reducing flow of packets through the network and saving energy by placing unused nodes in sleep mode. Jiang [11] proposed a novel localization approach where unknown nodes through their near anchor nodes to obtain their position. In order to reduce error during localization, a new means was used to approximate the distance between unknown nodes and anchor nodes when it is larger than node’s communication radius. Including this, self-adapting genetic algorithm is proposed to calculate the similar position of nodes, it makes the localization error much lower than the common method. Yetkin and Gungor [12] proposed a new Received Signal Strength Indicator (RSSI) based fingerprint technique which uses logical inferences. Here closed area was divided into the cells of 1 x 1 mt. The RSSI characteristics of each cell were recorded into a database in order to prepare a radio map. At real time, the RSSIs of anchor nodes received from base station were compared with radio map according to logical algorithms. In this scheme, the target localization was carried out mathematically. Wei Zhag [13] proposed a two-phase robust localization algorithm based on Consistency of Beacons in Grid. In the first-phase, a voting method based on the consistency of beacons in the grid is used to filter out part of the suspicious nodes. In the second-phase, it was adopted the loss function in M-estimation of Robust Statistics to obtain a robust solution with the remained nodes. Zhang and Hong Pei [14] explored a two-hop Collaborative Multilateral Localization Algorithm (CMLA). This algorithm was implemented through event-driven schemes. It is also introduced a new method which is used to estimate the distances between two hop nodes, applies anchor nodes within two hops to localize unknown nodes, and uses the minimum range error estimation to compute coordinates of unknown nodes. If any unknown node cannot be localized through two hop anchors nodes, it was localized by anchors and localized nodes within two hops. Chengpei [15] implemented a WSN localization method based on plant growth simulation algorithm (PGSA). This algorithm is a bionic random algorithm that characterizes the growth mechanism of plant phototropism. Based on simulation analysis, this algorithm (PGSA) is simple, fast convergence and robustness, which is more suitable for the large-scale environment. Long Cheng [16] presented a comprehensive analysis of these challenges: localization in non-line-of-sight, node selection criteria for localization in energyconstrained network, scheduling the sensor node to optimize the trade off between localization performance and energy consumption, cooperative node localization, and localization algorithm in heterogeneous network. Including this it was introduced that the evaluation criteria for localization in wireless sensor network. Oguejifor [17] implemented a localization system that uses a RSSI trilateration approach in a wireless sensor network. The system position estimation accuracy was also evaluated. Finally it was concluded that for the proposed system to work there must be the availability of at least three anchor nodes within the network and whenever anchor nodes broadcast packets containing their locations and other sensed parameters, the blind node within the broadcast range can always estimate its distance to the anchor nodes, and if peradventure the blind nodes receive packets from at least three anchors, the blind node can localize its position. Xiajoun Zhu [18] examined two candidate solutions developed from existing ideas, with one assuming that nodes can hear from each other if and only if they are within transmission range, and the other assuming closer nodes observe larger RSSI. Both candidate solutions do not work well in practice. After changing “closer” to “the closest” and “larger” to “the largest” in the second approach, it was found that the new assumption is quite reliable in practice. Rama and Parvadha [19] proposed a fuzzy logic-based restriction system suitable for remote sensor hubs that are portable in uproarious, savage situations. The constituent frameworks used fuzzy multi lateration and a grid prediction to process the area of a hub as a zone. The signal strength is thrown into bins which encode the imprecision. Laslo [20] presented WSN based fingerprinting localization method. The RSSI values of the communication links between the previously situated sensors and Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 34 – 42

Int J Adv Appl Sci

ISSN: 2252-8814



the mobile sensor were recorded in an indoor environment through the experiment. Using the recorded RSSI values a feed-forward type of neural network was trained. The result of the training is a neural network capable of performing indoor localization. The accuracy of the localization between the real and the calculated values was measured with Euclidean distance and demonstrated with the cumulative distribution function. Priti and Tyagi [21] proposed a technique called Multidimensional scaling which computes the position of nodes which are in the communication range of each other. This analysis technique find out the relative position of nodes with accuracy sufficient enough for most of the applications so as to solve the problem of recreation. Martin and Ramalakshmi [22] developed a localization system that carries high-location estimation accuracy at low cost. The system used spatiotemporal properties of well-controlled events in the network; light in this case, to obtain locations of sensor nodes. The system was to detect the multiple events in the network and to increase the area of the sensor field by increasing the number of nodes. By handling this kind of detection of multiple events in the network at once, mainly the time was saved. Sachin Deshpande [23] presented the methodology that gives a solution to compute the state parameters of the adversary target and tracks it and associate the same with the location in the periphery of wireless sensor networks. Nirmala [24] discussed a new technique that aims to localize all the sensor nodes in the network using trilateration, and a security protocol was used for providing confidentiality and authentication between anchor nodes and sensor nodes. Baihua and Guoli [25] proposed a new method, based on radial distance modulation, to detect and locate moving object from top view angle. This method has advantages of extracting information directly from the moving object characteristics of movement and spatial position, small computation, good robustness, convenient configuration, non-contact etc. It can locate the moving object with simple information after modulating and encoding the perception area of sensors. Dan and Daniel [26] proposed another anchor node localization technique that can be used when GPS devices cannot accomplish their mission or are considered to be too expensive. This novel technique was based on the fusion of video and compass data acquired by the anchor nodes and is especially suitable for video- or multimedia-based wireless sensor networks. Divya [27] proposed a mobility control scheme and we explored the impact of mobility over the performance of wireless sensor network. Two different protocols were used for the performance analysis of proposed mobility control scheme and the impact of this method over the selected protocols. It was analyzed the performance of the protocols on the basis of different parameters like Throughput, Packet Delivery Ratio, Routing Load and energy consumption. Jang Ping Sheu [28] proposed distributed localization scheme where each normal node gathers the necessary information via two-hop flooding and is thus scalable. Aside from this, each normal node uses a simplified approach and the proposed improved grid-scan algorithm to find the initial estimated locations of the normal node, thus reducing the computation cost. It also introduced a vector-based refinement scheme to correct the initial estimated location of the normal node, thus improving the accuracy of the estimated location. The nodes which are aware of their locations using special positioning devices are called anchor nodes or reference nodes. Other nodes that do not initially know their locations are called unknown nodes or sensor nodes. Generally, an unknown node estimates its location by range-based or range-free methods if three or more anchors are available in its 2-dimensional coverage field [29]. In all these better localization precision is achieved with the increased number of anchor nodes. The main problem with the increased number of anchors is that they are far more expensive than the rest of the sensors. The price of the whole network will increase even if only 10% of the nodes are anchor nodes. After the unknown/ stationary nodes have been localized, the anchors become useless. For this reason, it is necessary to consider an optimal number of mobile anchors to localize the sensor network [30]. The main idea of localization using a mobile anchor node is as follows a mobile anchor node traverses the sensor network by broadcasting anchor packets that contain the coordinates of the anchor node after the deployment of sensor network. Sensor nodes that receive anchor packets possibly will infer their distance from a mobile anchor node and use these measurements as constraints to construct and maintain position estimates. These methods have a common feature such as they use range-based approaches.Based on these analyses, localization using an optimal number of mobile anchor nodes would be more economy. In addition, it is necessary to consider the constraints in computing and memory power of sensors with an optimal number of mobile anchors for efficient localization in wireless sensor networks.

3. RESEARCH METHOD 3.1. System environment In the simulation, the considered WSN consists of two types of sensors including static sensor nodes and loge (n) number of mobile anchor nodes. Static sensor nodes are randomly deployed in a two-dimensional coordinate system. The locations of static sensors are unknown since they do not have GPS facility. The mobile anchor nodes are equipped with GPS receiver to determine their locations when they navigate over the sensing Trilateration based localization method using mobile anchor in wireless sensor networks (M. G. Kavitha)



ISSN: 2252-8814

region. The anchor nodes which are mobile in nature and sensors are able to receive messages from these anchors in the sensing region. 3.2. Beacon messages In the proposed model the anchor nodes periodically send message packets containing their position coordinates while navigating in the sensing area. These messages are called beacon messages. When an unknown sensor node receives at least three such message packets, it calculates its position coordinates using trilateration method. We assume that the distance between the mobile anchor nodes, any unknown node is estimated using RSSI technique. 3.3. Energy model The mobile anchor has sufficient initial energy for moving and broadcasting anchor packets during localization is the basic assumption we consider regarding energy level. Number of bits transmitted in a message and distance travelled affect the energy level of anchor nodes. 3.4. LUMAT Method 3.4.1. Network segmentation Assume the sensing region is divided into several hexagons. The mobile anchor nodes traverse the entire region in order to cover all sensor nodes. Since the sensing region is in irregular shape some of the regions may not be covered by hexagonal space. 3.4.2. Mobile anchor node The anchor nodes randomly traverse around the entire network, which periodically broadcast messages. The mobile anchor traverses the entire region at the speed V and broadcasts its current location (xi, yi) with an interval L and a communication range 𝑟 as depicted in Figure 2. The pseudo code for mobile anchor is described below. Input: {(x, y) – coordinates of anchor nodes, L-interval} Output: M-message Process: Set initial timer = 0 // broadcasting positions with an interval L If (timer % T = = 0) then (xi, yi) = GetPosition (xi, yi) Msg = MakeMessage (xi, yi) Broadcast (msg); End if 3.4.3. Unknown nodes Unknown nodes receive positions of the mobile anchor nodes continuously, and save the coordinates within a certain period of time “T”, i.e. the time that mobile anchor nodes complete all information broadcasting in the sensing region. If a node receives several messages at a time it considers the messages having higher signal strength and continues finding its location. Trilateration method is used to estimate positions (xesst, yest) of unknown nodes.

Figure 2. System environment with mobile anchor nodes Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 34 – 42

Int J Adv Appl Sci

ISSN: 2252-8814



The following is the pseudo code for unknown nodes Input: T – time period, M - message Output: (xest, yest) Process: Set number of received messages n = 0 Set clock initialization timer = 0 // receive messages continuously within a time period Of T While (timer! = T) { msg = ReceiveMessage () n++ RecordPosition (n, msg) } Calculate ( xest, yest) 3.4.4. Trilateration An example of the trilateration is shown in Figure 3. Unknown node 𝐷(𝑥, 𝑦) receives three anchor packets from the mobile anchor, namely, 𝐴(𝑥𝑎, 𝑦𝑎), 𝐵(𝑥𝑏, 𝑦𝑏), and 𝐶(𝑥𝑐, 𝑦𝑐). Distances between 𝐴, 𝐵, 𝐶, and 𝐷 are 𝑑𝑎, 𝑑𝑏, and 𝑑𝑐 respectively. Since the unknown node 𝐷 is within the regular triangle which is composed of 𝐴, 𝐵, and 𝐶, unknown node 𝐷 will calculate its location by using: ( − 𝑥𝑎)2 + (𝑦 − 𝑦𝑎)2 = 𝑑 𝑎 2 ( − 𝑥𝑏)2 + (𝑦 − 𝑦𝑏)2 = 𝑑 𝑏 2 ( − 𝑥𝑐)2 + (𝑦 − y𝑐)2 = 𝑑 𝑐 2

Figure 3. Trilateration based localization

4. RESULTS AND ANALYSIS 4.1. Evaluation criteria a. Localization Ratio Localization ratio is the ratio of the number of unknown nodes localized to the total number of unknown nodes. This metric also indicates the coverage degree of the movement path. Localization ratio is defined as Lratio = Nl / Nu Trilateration based localization method using mobile anchor in wireless sensor networks (M. G. Kavitha)



ISSN: 2252-8814

Where 𝑁𝑙 is the number of localizable unknown nodes and 𝑁𝑜 is the total number of unknown nodes. b.

Localization Accuracy The localization error of unknown node u is defined as eu =

(xu – au)2 + (yu – bu)2 + (zu – cu)2 R

Where (xu, yu, zu) are real coordinates of an unknown node u, (a u, bu, cu) are estimated coordinates of an unknown node u, and R is the communication range of sensor nodes. c.

Path Length To save energy consumption and time for localization, the path length of the mobile anchor node should be as short as possible.

Scalability Scalability means that the localization performance is independent of the unknown nodes density.

4.2. Simulations and analysis Hundred sensor nodes are randomly deployed in a 100m × 100m square region as shown in Figure 4. Each sensor can communicate with the mobile anchor nodes if distance between them is smaller than sensor range R. Figure 5 to Figure 7 show the localization error ratio, energy consumption of nodes and the coverage ratio localization.

Figure 4. Random distribution of nodes

Figure 5. Localization error ratio

Figure 6. Energy consumption of nodes

Figure 7. Coverage ratio during localization

E n e r g y

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 34 – 42

Int J Adv Appl Sci



ISSN: 2252-8814

The key metric for evaluating a localization technique is the accuracy of the location estimates versus the energy consumption and localization ratio. Localization accuracy mainly depends on increased density of anchors or number of the broadcasting messages, but the tradeoffs need to be made to determine appropriate deployment parameters. Table 1 presents the performance analysis.

Table 1. Performance analysis Metrics LUMAT SINGLE ANCHOR

Detection efficiency (%) 65-99

PDR (pkts) 23-93

Network L time (Secs) 456-987

End to end delay (msec) 0.32-0.11

Energy consumption % 11-19 %

Packet Integrity Vs Speed 12-25

47-68

18-55

256-687

0.562-0.315

29-32%

76-52

CONCLUSION In this paper we propose how to localize individual nodes in a network when the anchors are mobile in nature. An optimal number of anchors are considered for efficient localization of nodes in the network. Simulations and tests show that proposed localization method is energy efficient and accurate in nature. Also the ratio of coverage of nodes in the network is high. Simulation results show that our LUMAT algorithm significantly outperforms the localization method containing single anchor node in the network.Anchors should be designed dynamically or partially according to the observable environment or deployment situations to make full use of real-time information during localization. This is the future research issue in the area of mobile anchor assisted localization algorithm.

REFERENCES [1] [2] [3] [4] [5] [6] [7]

[8] [9] [10] [11] [12]

[13] [14] [15] [16] [17]

Z. Hu, D. Gu, Z. Song, and H. Li, “Localization in Wireless Sensor Networks Using a Mobile Anchor Node,” Proceedings of the 2008 IEEE/ASME, International Conference on Advanced Intelligent Mechatronics, 2008. K.F. Ssu, C.H. Ou, and H. C. Jiau, “Localization with mobile anchor points in wireless sensor networks,” IEEE Trans. on Vehicular Technology, vol. 54, no. 3, pp. 1187-1197, 2005. Yurong Xu, Yi Ouyang, Zhengyi Le, James Ford, and Fillia Makedon, “Mobile Anchor-free Localization for Wireless Sensor Networks,” International Conference on Advanced Intelligents, 2007. Jinfang Jiang, Guangjie Han, Huihui Xu, Lei Shu, and Mohsen Guizani, “LMAT: Localization with a Mobile Anchor node based on Trilateration in Wireless Sensor Networks,” IEEE Globecom proceedings, 2011. Sayyed Majid Mazinani and Fatemeh Farnia, “Localization in Wireless Sensor Network Using a Mobile Anchor in Obstacle Environment,” International Journal of Computer and Communication Engineering, vol. 2, no. 4, 2013. Kuo-Feng Ssu, Chia-Ho Ou, and Hewijin Christine Jiau, “Localization With Mobile Anchor Points in Wireless Sesnor Networks,” International Journal of Computer Communications, vol. 8, no. 6, 2015. Ming Chen, Xumin Xu, Shaohui Zhang, and Guofu Feng, “Energy Efficient Routing Protocol in Mobile-Sink Wireless Sensor Networks,” TELKOMNIKA (Telecommunication Computing Electronics and Control), vol. 10, no. 8, pp. 2056-2062, 2012. Madhu Patil and Chirag Sharma, “Energy Efficient WSN by Optimizing the Packet Failure in Network,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 7, no. 2, pp. 415-425, 2017. R. Khadim, M. Erritali, and A. Maaden, “Rang-Free Localization Schemes for Wireless Sensor Networks,” TELKOMNIKA (Telecommunication Computing Electronics and Control), vol. 16, no. 2, pp. 323-332, 2015. M. Ali Hussain, “Energy Efficient Intrusion Detection Scheme with Clustering for Wireless Sensor Networks,” TELKOMNIKA (Telecommunication Computing Electronics and Control), vol. 15, no. 1, pp. 128-141, 2015. N. Jiang, S. Jin, Y. Guo, and Y. He, “Localization of Wireless Sensor Network Based on Genetic Algorithm,” International Journal of Computer Communications, vol. 8, no. 6, pp. 825-837, 2013. Yetkin Tatar and Gungor Yildirim, “An Alternative Indoor Localization Technique Based on Fingerprint in Wireless Sensor Networks,” International Journal of Advanced Research in Computer and Communication Engineering, vol. 2, no. 2, pp.1288-1294, 2013. Wei Zhang, Wenqing Liu, Yunfang Chen, and Zeyu Ni, “Robust Secure Localization of WSN Based on Consistency of Beacons in Grid,” Journal of Computational Information Systems, vol. 10, no. 6, pp. 2283-2295, 2014. Shaoping Zhang and Hong Pei, “A Two-hop Collaborative Localization Algorithm for Wireless Sensor Networks,” TELKOMNIKA (Telecommunication Computing Electronics and Control), vol. 11, no.5, pp. 2432-2441, 2013. C. Tang, R.i Liu, and J. Ni, “A Novel Wireless Sensor Network Localization Approach: Localization based on Plant Growth Simulation algorithm” Elektronika IR Elektrotechnika, vol. 9, no. 8, pp. 97-100, 2013. Long Cheng, Chengdong Wu, Yunzhou Zhang, HaoWu, Mengxin Li, and Carsten Maple, “A Survey of Localization in Wireless Sensor Network,” International Journal of Distributed Sensor Networks, pp.1-13, 2012. O.S. Oguejiofor, V.N. Okorogu, Adewale Abe, B. Osuesu, “Outdoor Localization System Using RSSI Measurement of Wireless Sensor Network,” International Journal of Innovative Technology and Exploring Engineering, vol. 2, no. 2, pp. 1-6, 2013.

Trilateration based localization method using mobile anchor in wireless sensor networks (M. G. Kavitha)



ISSN: 2252-8814

[18] Xiaojun Zhu, Xiaobing Wu, and Guihai Chen, “Relative localization for wireless sensor networks with linear topology,” Elsevier, vol. 36, pp. 1581-1591, 2013. [19] M. Rama Prabha and R. Parvadha Devi, “Efficient Node Localization in Mobile Wireless Sensor Network,” International Journal of Advanced Research in Computer Science & Technology, vol. 2, no. 1, pp. 246-249, 2014. [20] Laslo Gogolak, Szilveszter Pletl, and Dragan Kukolj, “Neural Network-based Indoor Localization in WSN Environments,” Acta Polytechnica Hungarica, vol. 10, no. 6, pp. 221-235, 2013. [21] Priti Narwal and S.S. Tyagi, “Position Estimation using Localization Technique in Wireless Sensor Networks,” International Journal of Application or Innovation in Engineering & Management, vol. 2, no. 6, pp. 110-115, 2013. [22] M. Victor and K. Ramalakshmi, “Multiple Event-Driven Node Localization in Wireless Sensor Networks,” Int. J. of Advanced Research in Computer Engineering & Technology, vol. 2, no. 3, pp.1073-1077, 2013. [23] Sachin Deshpande, Umesh Kulkarni and Mritunjaykumar Ojha, “Target Tracking inWireless Sensor Network,” International Journal of Emerging Technology and Advanced Engineering, vol. 3, no. 9, pp. 177-181, 2013. [24] M. B. Nirmala, Nayana, and A .S. Manjunath, “Localization of Wireless Sensor Networks Using Robust Estimated Trust Evaluation Model”, Int. Journal of Scientific Engineering and Technology, vol. 2, no. 7, pp.729-732, 2013. [25] Baihua Shen and Guoli Wang, “Distributed Target Localization and Tracking with Wireless Pyroelectric Sensor Networks,” International Journal on Smart Sensing and Intelligent Systems, vol. 6, no. 4, pp. 1400-1418, 2013. [26] Dan Pescaru and Daniel-Ioan Curiac, “Anchor Node Localization for Wireless Sensor Networks Using Video and Compass Information Fusion,” Sensors, vol. 14, pp. 4211-4224, 2014. [27] Divya Bharti, Manjeet Behniwal, and Ajay Kumar Sharma, “Performance Analysis and Mobility Management in Wireless Sensor Network,” International Journal of Advanced Research in Computer Science and Software Engineering, vol. 3, no. 7, pp. 1333-1342, 2013. [28] Jang-Ping Sheu, Pei-Chun Chen, and Chih-Shun Hsu, “A Distributed Localization Scheme for Wireless Sensor Networks with Improved Grid-Scan and Vector-Based Refinement,” IEEE Transactions on Mobile Computing, vol. 7, no. 9, pp. 1110-1123, 2008. [29] K. Vinoth Kumar and S. Bhavani, “Localization based Optimized Energy Routing for Wireless Sensor Networks,” Middle East Journal of Scientific Research (MEJSR), vol. 23, no. 05, 2015. [30] K. Vinoth Kumar, T. Jayasankar, V. Srinivasan, and M. Prabhakaran, “EOMRP: Energy Optimized Multipath Routing Protocol for Wireless Sensor Networks,” International Journal of Printing, Packaging & Allied Sciences, vol. 4, no. 1, pp. 336-343, 2016.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 34 – 42

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 43~50 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp43-50



An immune memory and negative selection to visualizing clinical pathways from electronic health record data Mouna Berquedich1, Oulaid Kamach2, Malek Masmoudi3, Laurent Deshayes4 1,2

Laboratory of Innovative Technologies (LTI), Abdelmaleksaâdi University, Morocco 3 Laboratory of Industrial Engineering, Jean Monnet University, France 4 University Polytechnique de Benguérir, Morocco

Article Info

ABSTRACT

Article history:

Clinical pathways indicate the applicable treatment order of interventions. In this paper we propose a data-driven methodology to extract common clinical pathways from patient-centric Electronic Health Record data (EHR). The analysis of patient's, can lead to better regarding pathologies. The proposed algorithmic methodology consists to designing a system of control and analysis of patient records based on an analogy between the elements of the new EHRs and the biological immune systems. The detection of patient profiles ensured by biclustering Matrix. We rely on biological immunity to develop a set of models for structuring knowledge extracted from EHR and to make pathway analysis decisions. A specific analysis of the functional data leds to the detection of several types of patients who share the same EHR information. This methodology demonstrates its ability to simultaneously processing data, and is able to providing information for understanding and identifying the path of patients as well as predicting the path of future patients.

Received Jun 27, 2019 Revised Dec 19, 2019 Accepted Jan 11, 2020 Keywords: AIS EHR Hospital environment Immune memory Negative selection

This is an open access article under the CC BY-SA license.

Corresponding Author: Mouna Berquedich, Laboratory of Innovative Technologies (LTI), Abdelmaleksaâdi University, Tangier, Morocco. Email: berquedich.mouna@gmail.com

INTRODUCTION The use of EHRs that has developed around the world, more and more hospitals and health care providers are recognizing its benefits. However, it is difficult to identify the specific factors contributing to improved care. One of the main problems in this area is whether the information provided by the EHR is effective (in terms of better care and time) and whether it helps doctors in their decision-making. The results of several work on EHRs suggest that the use of EHRs improves medical decision-making in terms of accuracy of diagnosis and admission of correct decisions, as well as better quality of care. This study contributes to medical decision-making in that the results show how access to EHR can improve patient care and save time and money. the objective of this work is to rely on the EHR to increase the cooperation and willingness of medical personnel to adopt a computer system by demonstrating its contribution to the correction of medical diagnostics. The system developed is based on a set immune concepts and mechanisms, such as the negative selection algorithm used to monitor, and an immune memory algorithm used to select the appropriate pathway detection strategy to respond to detected disturbances. The proposed system maintains the performance of a clinical decision support system at a high level. The combination of negative selection mechanisms and immune memory gives the system the ability to recognize disturbances Journal homepage: http://ijaas.iaescore.com



ISSN: 2252-8814

and select appropriate decisions. In order to improve the system described in this article in, other immune concepts may be used. These concepts inspired by the biological immune system, such as the theory of danger, can be studied to take into account the influence of a bad diagnosis on the prediction of the paths of other patients. The patient pathway could also refer to the succession of steps in the handling of a patient within a hospital [1]. This may refer to a sequence of procedures (e.g. clinical examination, laboratory dosage, biopsy, and then surgery), a sequence of clinical stages (e.g. inflammatory, proliferative, and maturation phases), or a sequence of medical units (e.g., emergency unit, surgery, intensive care, conventional hospitalization, rehabilitation care, and then housing unit) [1]. In this context, the unprecedented availability of hospitals data gives the opportunity to improve decision-making and to discover best practices for healthcare delivery [2]. This article presents a case study of EHRs—the study of the management process of the hospital medical EHR and the construction of futur Pathaways of patients in order to identify patient’s profiles from the EHR data of the already occurred patients’ pathways [3-7]. Electronic Health Records (EHR) is a digital collection of patient health information. EHR is increasingly being implemented in many developing countries. These records can be shared through well connected network across different health care settings. EHR includes demographic and personal statistics like age, weight, and billing information as well as vital signs, family history, medication and allergies history, immunization status, laboratory test results, and radiology images. EHR systems are designed in such a way to reduce paper medical records by storing data accurately and legible in a digital format. However digitally health records reduce the risk of data replication as file can be shared across the different health care systems and can be easy updated which reduce the risk of lost paperwork. EHR programs directly benefit the physicians, patients, and obviously the hospital management authority. EHR system used for population based studies and effective when extracting medical to predict possible trends in healthcare system [8]. EHR is a great tool to manage lengthy and labor-intensive paperwork more efficiently and thereby significantly reduce the cost of transcription, re-filling, and storage. EHR enables patient management with enhanced and accurate reimbursement coding. Since the software has all patient-related information it significantly reduces the occurrence of a medical error and also helps in the improvement of patient health with better management of the diseases. Here we discuss five important benefits of EMR vs. Paper Medical Records. (a) Costs: to start the EHR the initial costs is higher due to large and digitally setting of IT network but the costs over time will decrease significantly. While manually storage of paper records require more personnel to manage and maintain paper files, accesses and organize countless documents which increase cost substantially over the time. EHR can save man power, time and physical storage space which reduced the cost in long run. (b) Storage: electronic health records can be stored in a secure cloud, providing easier access by those who need them, however paper medical records required large warehouses for storage. Paper records are not only taken up space, but they are not environmentally friendly and tend to decay, when handled by many individuals over time, which increased the cost of storage. (c) Security: security is a great concern for both paper and electronic storage system; both are equally susceptible to security threats. If a facility stores records electronically without proper and effective security systems they are vulnerable to access by unauthorized individuals which can be misuse the information. If records are stored in paper form, they can be lost or damage or stolen due to human error. Natural disaster such as a fire or flood also plays an important role in the concern the security of health records. (d) Access: accessibility of electronic health records take clear advantage over paper health records. Digital health records allows healthcare professionals to access the information instantly, whenever and wherever they need almost, which makes healthcare professionals more efficient, however paper medical records to be shared with healthcare professional required physically provided to them or scanned and sent via email, it is a time-consuming which increase the cost. (e) Readability and Accuracy: electronic health records are often written with the use of standardized abbreviations which make them more accurate and readable across the globe which decreased the chance for confusion however handwritten paper medical records may be poorly legible, which can contribute to medical errors. Paper medical records provide insufficient space for healthcare professionals to write all necessary information [9]. In terms of appointment management, EHR performs the excellent task. EHR improved medical practice management through integrated scheduling systems that link appointments directly to progress notes, automate coding, and manage claims [10]. This platform smoothly handles queries about the patient condition and manages graphs, specific to each and every patient. EHR enhanced communication with other multi-disciplinary physicians, laboratories, and between different hospitals which enables faster patient service. Since it’s a digital platform with online connectivity doctors can access the patient information anytime anywhere and assigned task to support providers which includes labs, and other physicians [11]. Follow checkups are often an integral part of better patient care; with EHR, automated checkups are easily scheduled by the electronic program. Since the program is integrated with test reports and images, timely Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 43 – 50

Int J Adv Appl Sci

ISSN: 2252-8814



access is readily enabled. Moreover, such test can be ordered through this multipurpose EHR system. Moreover, such system prevents unnecessary duplication of medical testing. The best part of such electronic records is that they are integrated with national and international disease database and registries. Thus, it helps the physicians track the epidemiological status of the current disease under treatment and be prepared for emergencies. 2. LITERATURE REVIEW 2.1. Artificial immune systems The Artificial Immune System field has been inspired from natural immune system of several species. Ambitiously, to develop systems that operate in environments similar to constraints faced by the natural immune system. De Castro and Timmis defines the AIS as “the adaptive systems, inspired by the theories of the immunology, as well as the functions, the principles and the immune models, in order to be applied to the resolution of problems” [12]. The immunity is subdivided into two distinct systems: innate immune system and adaptive immune system. The adaptive immune system has three principal processes [13]: negative selection, clonal selection and immune network. Whereas, Natural Dendritic Cells are the link between the innate and adaptive immune system. 2.2. Clonal selection There are many algorithms based on clonal selection in the literature, most of which have been applied to optimization problems (e.g., CLONALG [14] and opt-IA is used in [15], and the algorithm to B cells in [15]), and multi-objective optimization [16]. From a computational point of view, the idea of clonal selection leads to algorithms that iteratively improve the possible solutions to a given problem through a process of cloning, mutation and selection (in a way similar to genetic algorithms [15]). In this section, we describe an algorithm based on clonal selection that has been used as a basis for theoretical studies [16]. 2.3. Immune network The immune network theory is originally proposed by Jerne [17]. An artificial immune network is a bio-inspired informatics model that uses the ideas/concepts of the immune network theory, mainly, the interaction between B cells and the cloning process. It receives an antigen as entry and sends back an immunized network compound of the B cells that are adjusting in between. The immune network process is almost the same as clonal selection, except that there exists a mechanism of deletion that destroys the cells having a certain inception of affinity amidst. 2.4. Algorithm of negative selection The basic idea of a negative selection algorithm is to generate a number of detectors in the complementary set N, then applying these detectors to classify the new data as auto or non-auto [18]. Negative selection algorithms have been very widely used in aquatic invasive species research and have undergone many improvements over the years [19]. Clonal selection algorithms are mostly used as optimizing algorithms. They use fewer classifications. Thus, clonal selection principles first appeared in 1959 [20]. Artificial immune systems algorithms used for classification are considered as classifiers, since they combine the output of many simple classifiers all together. 3.

AIS METHODS The Two types of cells involved to recognize presenting pathogens are; T-cells and B-cells. The population of these present cells in the bloodstream is responsible for recognizing and destructing the pathogens. The population acts collectively. It is capable to identify new pathogens through two training methods: negative selection and clonal selection. Through the process of negative selection, the NIS is able to protect the host organism tissues from being attacked by its own immune system. Some cells generate detectors that recognize proteins, which are present on the surfaces of cells. The detectors are called “antibodies”. They are randomly created. Before the cells become fully mature, they are “tested” in the thymus. The thymus, an organ located behind the sternum, is able to destroy any immature cells that identify the tissues of the organism as “non-self” [21]. The process of negative selection maps therefore the negative space of a given class such as given examples of the “self” class. The negative selection algorithm first appeared in 1994 [22]. Using the clonal selection, the NIS is apt to adjust itself to provide the most efficient response against pathogen attacks. Clonal selection happens when a cell detector finds already seen pathogen in the organism, it clones itself then to start the immune response. The cloning process, however, introduces small variations in the pattern that the cell detector recognizes. The number of the clones created by a cell detector is proportional to the new pathogen cell “affinity". It is a measured manner to detect An immune memory and negative selection to visualizing clinical pathways from… (Mouna Berquedich)



ISSN: 2252-8814

to which extend the cell matches the pathogen. The amount of variation allowed in the clones is negatively proportional to the affinity; the cells with the most affinities are mutated less. The clonal selection algorithms are similar to the natural selection systems. And the clonal selection algorithm is therefore similar to the genetic algorithms based on the natural selection [23]. Nevertheless, clonal selection algorithms have less parameters than genetic algorithms, and potentially, they do not require complex functioning operations. Clonal selection algorithms are mostly used as optimizing algorithms. There have been a few of them used for classification. The clonal selection principle first appeared in 1959 [23]. Artificial immune systems algorithms applied for classifications are considered to being classifiers; they combine the output of many simple classifiers. 3.1. System overview The main objective of our work is to develop a vital supporting tool for hospital decision-makers to strengthen the quality of their decisions face of the massive flow of patients. The fundamental idea is to detect traces in the database, and to help executives by identifying bad scenarios utilizing AIS techniques, especially negative and clonal selection. The analogy between the principle of the natural immune system and the problem as proposed in Table 1 has prompted us to develop our system. Table 1. Analogy Between the Natural Immune System Principle and the Developed System of EHR Management Natural Immune System Body Self Infected Cell Non-self (antigen) antibody Lymphocyte (B) Affinity Memory Cells Response Strategy

Artificial Immune System applied in our Hospital emergency context HER Normal Pathway of patient Disturbed Pathway False Pathway Control Decisions Combination of control decision for detected disturbance Adequacy between the correction actions and disturbance. Data base Immune Memory Based Algorithm

3.2. Self-cell representation Self-cells represent normal situations of EHR. In this paper, we suggest a model to represent and structure knowledge related to normal situations of EHR. The model includes five attributes as presented in (1). SE= {D,Prj,Pj,Mj,Dj}

(1)

Dj: Date Prj: Purpose Pj: Procedure Mj: Medication Dgj: Diagnosis 3.3. Antigen represntation A disturbance is any kind of event that affects a pathway of patient. We characterize a disturbance as a vector of patient i with 5 attributes describing the affected pathway. Vi { Date Purpose Procedure Medication Diagnosis }

(2)

e lk, (k = 1, . . ., kl, l = 1,. . ., l), represents a set of kl events of l specific types, referred to as ‘scenario’ hereafter, that occurs during a patient’s medical visit. In Table 1, the "Office" event corresponds to the type of contact, the "CKD Stage 3" event to the type of diagnosis, the "Diuretics" event to the type of medication, and the "Renal" type of ultrasound; procedure. For example, in Table 1, the record of patient 1 can be transformed using the scenario shown in Table 2. The procedure, treatment and diagnostic scenarios are named respectively Pj, Mj and Dj , j ∈ Z +. Our goal is to identify common sequences from data that can constitute clinical pathways. To do this effectively, we introduce an element called immune memory and negative selection to represent unique visitor content. For each patient we will have a unique combination: Purpose of visit, procedure, medication and diagnosis. Negative selection to reduce multidimensional visit records so that they can be represented as Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 43 – 50

Int J Adv Appl Sci



ISSN: 2252-8814

a sequence of visits ordered by date of visit. Each patient has one and only one sequence, starting with the first visit recorded in the EHR and ending with the last visit. The description is illustrated in Table 3. Table 2. Description of visit (A) Patient 1 1 1 1

Date xx/yy/2012 xx/yy/2013 xx/yy/2015 xx/yy/2018

Purpose P1 P2 P3 P4

Medication M1 M2 M3 M4

Procedure N/A P1 N/A N/A

Diagnosis D1 D2 D1 D3

Table 3. Description of visit (B) Patient 1 1 1 1

Visit Date 24/09/2012 5/2/2013 3/1/2015 3/6/2018

Visit Purpose P1 P2 P3 P4

Procedure N/A Renal ultrasound N/A N/A

Medication ACE inibitors ACE inhibitors dueritic ACE inhibitors, dueritic statin ACE Inhibitors Dureritic statin

Diagnosis CKD stage 4, hypertension AKI, CKD Stage 4hypert CKDS Stage 4; hypertension AKI, CKD Stage 5, hypertension

The developed Immune Memory based Algorithm (IMA) works according to the following steps: a. Representation of B cells The control strategies are represented by the "B cells" that allow the recognition process and neutralize the antigen detected in the sphere concerned. They aggregate with one or more decision controls (antibodies) that react each time the disturbance occurs. Therefore, the system must create a B cell identical to each detected antigen. The (1) illustrates the definition of independent B-cells that receptors are a prerequisite that has the structure similar to that of the antigen that has been described in (3). The set of receptors, epitopes and antibodies is equal to the B cells. According to the model suggested in, a control decision can have the value 0 or 1. Whether the antibody is operated or activated, the similar epitope indicates the value 1 and the value 0 is assigned to it otherwise. Epitopes are the activators of antibodies in these cases. b.

Step 1: Learning This step aims to produce sets of non-self cells using periodic comparisons between normal situations (self cells) and the state of the new EHR. We highlight matching ratios to quantify existing distances in given situations, using data collected from the facts. These measures are designated by SE and the elements of the set S with respect to testing their similarities. The matching rate is calculated mathematically by adapting (3). Mat (Si,SE) =

×∑ 𝛼𝑖

(3)

The «Mat (Si, SE)» represents the corresponding percentage (%). The "Si" indicates a normal situation among the set "S". SE determines a situation. Attribute values are extracted from the database. αi is calculated using the following method (4). αi= 1 𝑖𝑓 𝑆 = 𝑆 0 𝑒𝑙𝑠𝑒

(4)

The "SEj" is the "j" th attribute of an "SE" situation captured from the database (look (1)). The «𝑆𝑖𝑗» is the "j" th attribute of an "SE" situation captured from the database (look (1)). The «𝑆𝑖𝑗» is the "j" th attribute of the "i" th situation extracted from the set S. When the adequacy rate is lower than the coverage already fixed "1" e; therefore, the situation is classified as abnormal and will then be added to the R set of patterns. To close this process, the set R, this includes various types of deviant pathway, displays the most significant abnormal pathway. In fact, it is used in the next steps to identify the disturbances that have occurred. Figure 1 draws the components of the set R.

An immune memory and negative selection to visualizing clinical pathways from… (Mouna Berquedich)



ISSN: 2252-8814

Figure 1. The suggested Negative Selection Algorithm (NSA) c.

Step 1: Monitoring The given R which represents all the patterns is constructed in the step indicated above. The PS is designed to progressively read the values from the databases and measure the similarities with the situation of abnormalities of the R-set, non-auto-cellular cells. The (4) can be used as a determinant of the matching rate which groups all the abnormal situations of the set R with a current situation. Whenever the actual situation is the same as that already coded in the set R, that is, the matching rate is much higher than the fixed interval, it follows that perturbation is identified and a specific antigen created. The reaction and response will then be activated by the PS. 4.

RESULTS AND ANALYSIS This section presents the implementation of the EHR management system and discusses the results obtained. The proposed decision support system is implemented with the JAVA programming language. To evaluate the system, a hospital database with a history of four years with more than 100,000 EHR patients. We have simulated the different cases presented below. To construct the S set, we used a real patient records database (EHR). The illustrations of an example sample EHR on our system is shown as: STARTING PHASE > CLEANING PHASE >> EHR ENTRY VECTORS: [ EHR [patientId=-1, date=Tue Sep 25 21:59:42 WEST 2018, purpose=P10, procedure=P1, medication=M1, diagnosis=D5], EHR [patientId=-1, date=Tue Sep 25 21:59:42 WEST 2018, purpose=P3, procedure=P2, medication=M8, diagnosis=D5], EHR [patientId=-1, date=Tue Sep 25 21:59:42 WEST 2018, purpose=P2, procedure=P1, medication=M2, diagnosis=D5], EHR [patientId=-1, date=Tue Sep 25 21:59:42 WEST 2018, purpose=P2, procedure=P5, medication=M5, diagnosis=D2], EHR [patientId=-1, date=Tue Sep 25 21:59:42 WEST 2018, purpose=P10, procedure=P9, medication=M3, diagnosis=D10] ] >> OPTIMAL SOLUTIONS (DISTANCE ORDER): {80.0%= Patient [id=1, nom=N_ABCDEFGHIJ, prenom=P_R, adresse=ADRESSE, dateNaissance=2018-0924 22:21:33.0, ehrs=[EHR [id=1, patientId=1, date=2018-09-24 22:21:33.0, purpose=P10, procedure=P6, medication=M1, diagnosis=D5], EHR [id=2, patientId=1, date=2018-09-24 22:21:33.0, purpose=P3, procedure=P2, medication=M8, diagnosis=D5], EHR [id=3, patientId=1, date=2018-09-24 22:21:33.0, purpose=P2, procedure=P10, medication=M2, diagnosis=D7], EHR [id=4, patientId=1, date=2018-09-24 22:21:33.0, purpose=P2, procedure=P6, medication=M5, diagnosis=D2], EHR [id=5, patientId=1, date=2018-09-24 22:21:33.0, purpose=P10, procedure=P1, medication=M3, diagnosis=D10], Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 43 – 50

Int J Adv Appl Sci

ISSN: 2252-8814



EHR [id=6, patientId=1, date=2018-09-24 22:21:33.0, purpose=P1, procedure=P3, medication=M6, diagnosis=D5], EHR [id=7, patientId=1, date=2018-09-24 22:21:33.0, purpose=P5, procedure=P4, medication=M8, diagnosis=D9]]] , 40.0%= > END PHASE > LOAD TIME : 3971 ms The results above show the phase of the regrouping of the relevant information in the form of a vector of five components, thereafter an analysis is conducted according to two immune processes to communicate to the hospital decision maker the path closest to the path of the patient entered. The execution phase included a grouping and vector standardization compatible with the input patient information. Subsequently two immune algorithms as previously explained the negative selection and immune memory for the display of the optimal solution that responds to the patient input. The goal has been achieved, by offering the hospital decision maker a tool for detecting the patient's journey, in broad and unstructured information. The most important is the fact of constantly feeding the database R, containing the erroneous solutions in order to avoid the medical errors of the old patients during their medical journey to the recent patients who share the same pathology and symptoms with them. Figure 2 shows the result of HER selection. As illustrated in Figure 2, we realized the efficient and effective decision support system, allowing the filtering and the analysis of the data of the EHRs of the different patients while adopting the immune principles of the memory cells and of negative selection.

Figure 2. Results of EHR selection by affinity performance 5.

CONCLUSION We are designing a patient record analysis system, based on an analogy between the elements of the new HRTs and the biological immune systems. Patient profiles are detected by the memory cells. We rely on biological immunity to design a set of models that can be used to structure knowledge about EHRs and pathway analysis decisions. A specific analysis of functional data, led to the detection of several types of patients, who share the same information on their EHR. This methodology demonstrates its ability to simultaneously process data. It is able to provide information for the understanding and identification of patients' pathways as well as for predicting the path of future patients. REFERENCES [1] [2] [3]

J. E. Barr and J. Cuzzell, “Wound care clinical pathway: a conceptual model,” Ostomy Wound Manage, Vol. 42, No. 7, pp.18–24, 1996. B. T. Denton, Handbook of healthcare operations management. New York: Springer, 2013. A. Kellermann, “Crisis in the Emergency Department,” New England Journal of Medicine, Vol. 355, No. 13, pp. 1300–1303, 2006.

An immune memory and negative selection to visualizing clinical pathways from… (Mouna Berquedich)

50 [4] [5] [6] [7] [8] [9]

[10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23]



ISSN: 2252-8814

F. Harrou, F. Kadri, S. Chaabane, C. Tahon, and Y. Sun, “Improved principal component analysis for anomaly detection: Application to an emergency department,” Computers & Industrial Engineering, Vol. 88, pp. 63–77, 2015. F. Kadri, C. Pach, S. Chaabane, T. Berger, D. Trentesaux, C. Tahon, and Y. Sallez, “Modelling and management of strain situations in hospital systems using an orca approach,” in Proceedings of 2013 International Conference on Industrial Engineering and Systems Management (IESM). IEEE, pp. 1–9, 2013. F. Kadri, F. Harrou, S. Chaabane, and C. Tahon, “Time series modelling and forecasting of emergency department overcrowding,” Journal of medical systems, Vol. 38, No. 9, pp. 1–20, 2014. F. Harrou, Y. Sun, F. Kadri, S. Chaabane, and C. Tahon, “Early detection of abnormal patient arrivals at hospital Emergency department,” In 2015 International Conference on Industrial Engineering and Systems Management (IESM), IEEE, pp. 221-227, 2015. F. Kadri, S. Chaabane, and C. Tahon, “A simulation-based decision support system to prevent and predict strain situations in emergency department systems,” Simulation Modelling Practice and Theory, Vol. 42, pp. 32–52, 2014. P. Carey, G. Cuthbert, R. Dang, B. Greystoke, A. McGregor, R. Oakes, and J. Wallis, “The North of England haemato-oncology diagnostic service (NEHODS): A more devolved and inclusive approach to integrated reporting facilited by an IT system (Haemosys) networked to local information management systems (LIMS) in all participating regional hospitals,” British Journal Of Haematology, Vol. 173, No. 1, pp. 43-43, 2016. J. V. de Carvalho, Á. Rocha, and J. Vasconcelos, “Towards an Encompassing Maturity Model for the Management of Hospital Information Systems,” Journal of Medical Systems, Vol. 39, No. 9, p. 99, 2015. C. Virenque, Large influx of injured people in hospital. Hôpital Purpan, TSA 40031, 31059 Toulouse cedex 09, France Elsevier, pp. 712-715, 2016. L. N. De Castro and J. Timmis, “Artificial immune systems: a novel paradigm to pattern recognition,” Artificial Neural Networks in Pattern Recognition, Vol. 1, pp. 67-84, 2002. J. Timmis, A. Hone, T. Stibor, and E. Clark, “Theoretical Advances in Artificial Immune Systems,” Theoretical Computer Science, Vol. 403, No. 1, pp. 11-32, 2008. U. Aickelin and D. Das Gupta, “Artificial immune systems,” Search Methodologies, pp. 375-399, 2005. L. N. De Castro, and F. J. Von Zuben, “The Clonal Selection Algorithm with Engineering Applications,” Paper presented at The Workshop on Artificial Immune Systems and Their Applications, Las Vegas, USA, 2000. L. N. De Castro and F. J. Von Zuben, “Learning and Optimization Using the Clonal Selection Principle,” IEEE Transactions on Evolutionary Computation, Vol. 6, No. 3, pp. 239-251, 2002. N. K. Jerne, “Towards a Network Theory of the Immune System,” Ann Immunol, Vol. 125, pp. 373-389, 1974. J. Greensmith, U. Aickelin, and G. Tedesco, “Information Fusion for Anomaly Detection with the Dendritic Cell Algorithm,” Information Fusion Journal, Vol. 11, No. 1, pp. 21-34, 2010. J. Greensmith, U. Aickelin, and S. Cayzer, “Introducing Dendritic Cells as a Novel Immune-Inspired Algorithm for Anomaly Detection,” International Conference on Artificial Immune Systems, Springer, Berlin, Heidelberg, pp.153-167, 2005. M. Burnet, The Clonal Selection Theory of Acquired Immunity. Nashville, Vanderbilt University Press, 1959. B. Schmidt, A. Al-Fuqaha, A. Gupta, and D. Kountanis, “Optimizing an artificial immune system algorithm in support of flow-Based internet traffic classification,” Applied Soft Computing, Vol. 54, pp. 1-22, 2017. S. Forrest, A. Perelson, L. Allen, and R. Cherukuri, “Self-nonself discrimination in a computer,” Research in Security and Privacy, 1994. Proc. 1994 IEEE Computer Society Symposium, pp. 202-212, 1994. S. Mnif, S. Elkosantini, S. Darmoul, and L. Ben Said, “An immune multi-agent based decision support system for the control of public transportation systems,” International Conference on Practical Applications of Agents and Multi-Agent Systems, Springer, pp. 187-198, 2016.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 43 – 50

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 51~61 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp51-61



A computer vision-based weed control system for low-land rice precision farming O.M. Olaniyi1, E. Daniya2, J.G. Kolo3, J.A. Bala4, A. E. Olanrewaju5 1,5

Department of Computer Engineering, Federal University of Technology, Nigeria 2 Department of Crop Production, Federal University of Technology, Nigeria 3 Department of Electrical and Electronics Engineering, Federal University of Technology, Nigeria 4Department of Mechatronics Engineering, Federal University of Technology, Nigeria

Article Info

ABSTRACT

Article history:

Agricultural sector is one of the economic pillars of developing nations, because it provides means of boosting gross domestic profit. However, weeds pose a threat to food crop by competing with it for nutrients and undermining the profit to be made from it. The treatment of these weeds is necessary, but at minimal impact on the actual food crop. Herbicide usage is one major means of weed control, owning to the expensive and labourintensive nature of hand weeding. Recently, the need for site specific spraying has been on the rise because of health concerns which have been raised on the effect of herbicides on food crops and the effect on the environment. Most research on the field focuses on accurately identifying the weeds whilst neglecting the weed control. In this research, we apply fuzzy logic-based expert system to control how herbicide is sprayed on lowland rice in order to reduce excessive herbicide usage. The system supplies the control with weed density (Box size) and confidence level. The values of both are then passed to the fuzzy logic control for spray decision. The Sugeno as well as Mamdani models were tested using generated values for detected weed box size and confidence levels of the computer vision. The mean absolute error obtained was 0.9 for both, and 0.3 and 0.2 respectively, for the mean square error. The error shows how accurate the system can be and with low error value, it shows that the system implementation is capable of providing control for spraying of herbicides which in turn will yield more returns for low-land rice farmers.

Received Jun 28, 2019 Revised Sep 15, 2019 Accepted Jan 7, 2020 Keywords: Agriculture Computer vision Fuzzy inference system Low-land rice Precision farming

This is an open access article under the CC BY-SA license.

Corresponding Author: Olaniyi O. Mikail, Department of Computer Engineering, Federal University of Technology, Minna, P.M.B 65, Gidan Kwano, Minna, Niger State, Nigeria. Email: mikail.olaniyi@futminna.edu.ng

INTRODUCTION Precision farming is a cohesive production and information-based farming system with a goal to promote efficiency, boost production and lucrativeness of the farm production activities while avoiding the adverse effects of excessive chemical usage on the environment or inadequate application of input [1, 2]. Precision farming has been affirmed as the solution to sustainable agriculture with focus on production boost [3]. Agricultural sector is an important sector of any economy as a mean of providing food supplies to meet the populace demand.

Journal homepage: http://ijaas.iaescore.com



ISSN: 2252-8814

Rice (Oryza sativa), a plant species, is a seed of grass species grown as an annual plant, which is fast becoming a main food crop because of high patronage by a lot of people. It is consumed by more than half of world’s population and provides twenty percent of the calories consumed worldwide by humans. It also a major staple food of most elite homes in Nigeria [4]. Rice production goes through three stages, which are the vegetative phase, reproductive phase and the maturity phase. The difference in rice varieties of the world is the vegetative phase. Rice is farmed in about 1.7 million hectares of the estimated 4.6 million to 4.9 million hectares’ potential land for its production. The production environment for rice in Nigeria are the rain-fed lowland, rain-fed upland, irrigated lowland, deep water/floating and mangrove swamp [5]. Weeds are some of the major problems facing rice production in Africa [6]. Weed control mechanism be applied within 40 to 50 days of planting, else the control might prove difficult [7]. Likewise, author in [8] stated that weed control has high chances of improving production. Weeds cause increase in production cost, reduction in profit and contamination of food crops [9, 10] Other constraints on rice production are labour shortages, illiteracy, ignorance, poor milling system, pest infestation, poor drainage, inputs and credits [5, 11]. Due to increase in civility and knowledge of weed, means of weed control have been sought after by researchers to remove the notorious pest at minimal damage to the plant. Cultural methods, chemical methods and automated methods are the major means of weed control. The cultural method of weed control consists of maintaining clean reaper, mulching, fire clearance, early flooding, bush fallowing, hand weeding and shifting cultivation [12-16]. This method suffers from high cost and huge labour intensiveness. Herbicide application is seen as an important substitute to hand weeding. But over application of herbicides can lead to losses at harvest, environmental damage, high cost of production and building of resistance to the herbicide [7, 8, 17]. Some of these herbicides even end up on food crop and the soil without reaching the weeds [18] The spraying on food crop is seen as threat to safety of food consumed, this therefore breeds a need for comprehensive control system for management of weed. However, author in [19] proposed an image processing method based on complex pre-processing steps to accurately identify weeds in farm lands but provide no removal techniques for the weeds identified. In the same vein, [20] proposed a robotic system that classifies weeds based on visual texture. The system uses knives for removing the weeds and this might cause damage to the plant. The designed system also consumes a lot of power. Consequently, Pulido et al., in [21] developed a system that uses patterns from texture to identify weeds from vegetables using Support Vector Machine (SVM) classifier. Features space was calculated from grey level co-occurrence matrix (GLCM). The system obtained 90% for sensitivity, specificity and precision but provides no removal techniques for the weeds. In addition, author in [7] designed a weed detection system for rice in order to avoid uniform spraying of rice farm, the system used SVM for the classification of the weed and uses blur detector for weed detection. The system had an accuracy of 68.95% with blur and an accuracy of 76.16% when the blur was removed. However, no means of weed control was proposed or implemented. There are other existing methods that detect weeds in wheat, soya beans and maize with less accuracy in real life field application in literature. From the reviewed literatures, it is evident that implementation of weed control systems to solve the problem of excessive herbicide spraying has not gained much attention in recent times. Hence, this paper proposes a fuzzy logic-based expert control system for herbicide spraying control to help address the problem of uniform spraying and excessive herbicide usage. The remaining part of this paper is organised as follows: section 2 presents the fuzzy logic control system, section 3 presents the system design, section four presents the results obtained and section 5 presents the conclusion and recommendation for future works. 2. FUZZY LOGIC CONTROL SYSTEM 2.1. Fuzzy inference system Explaining The fuzzy logic is based on the work of Jan Lukasiewicz (1878-1956) and the term was coined out and developed by Dr. Lotfi Zadeh in 1965 [22]. Fuzzy logic is a pattern of reasoning which sets out to mimic the reasoning abilities of humans. It involves the use of continuous values from 0 to 1. These values can also be taken as true and false so that decisions can be made even with incomplete or uncertain data. It works on the levels of different possibilities in order to achieve an output. Fuzzy logic helps to give a reasonable solution to uncertain engineering problems. There are basically two models of fuzzy inference system which are the Mamdani model created by Mamdani and Assilian in 1975 and Takagi Sugeno Kang model created in 1985 [23]. Sari et al. described the four parts of the fuzzy logic inference system as: Fuzzification module which is the part first part that transforms the input of the system into (crisp) fuzzy set [24]. The knowledge Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 51 – 61

Int J Adv Appl Sci



ISSN: 2252-8814

base forms the second, it holds the if -then rules provided. The inference engine which is the third simulates human decisions by taking decisions based on the knowledge based created. The fourth is the defuzzification module that transforms the fuzzy set into crisp value in the Mamdani FIS. In the Sugeno FIS, weighted average or weighted sum is used for obtaining crisp value for the output. Figure 1 shows the block diagram that shows the four basic parts of fuzzy inference system.

Figure 1. Block diagram showing the different parts of FIS There are five defuzzification methods for Mamdani which are, smallest of maximum, centroid of area, largest of maximum, bisector of area and mean of maximum. Table 1 shows the defuzzification methods used in Mamdani and Sugeno, together with their mathematical formulas. Finally, the membership function shows graphically the relationship between the elements. The membership function works on fuzzy sets of the variable. There are different types of membership functions which are used in fuzzy inference systems and examples are singleton, Gaussian, triangular and trapezoidal. For this research both the Sugeno FIS and the Mamdani FIS were simulated for the control system, but the Sugeno was selected for implementation because of the flexibility, adaptive nature and its computational efficiency. Table 1. Methods of obtaining crisp output for Mamdani and Sugeno [23-25] Serial number

FIS model

Defuzzification method

Mamdani

Smallest of maximum

Mamdani

Centroid of area

Mamdani

Largest of maximum

Meaning It uses the smallest value which gives the maximum membership degree of the fuzzy set to generate the crisp output It generates the centroid of the area formed by the fuzzy set. It uses the value to calculate crisp output Just like SOM but it uses the largest value which gives the maximum membership degree to yield the final crisp output

Formula 𝑦

= min{𝑦/ 𝜇𝑩 (𝑦) =max (𝜇𝑩 (𝑦)} 𝑦. 𝜇𝑩 (𝑦)𝑑𝑦 𝜇𝑩(𝑦)𝑑𝑦

𝑦 = max{𝑦/ 𝜇𝑩 (𝑦) =max (𝜇𝑩 (𝑦)} 𝜇𝑩 (𝑦)𝑑𝑦

Mamdani

Bisector of area

Mamdani

Mean of maximum

Sugeno

Weighted average

Sugeno

Weighted sum

Bisector of area generates a vertical line which splits the fuzzy set into two equal areas. The vertical line corresponds to the output generated. This takes the mean of the maximum as the crisp output It uses the weighted output of the rule to generate the final output It reduces the computation by using the sum of the weighted output of the rule.

∝

𝜇𝑩 (𝑦)𝑑𝑦

∝ = min{𝑣/𝑣 ∈ 𝑉 ) 𝛽 = max {𝑣/𝑣 ∈ 𝑉} 𝑦 + 𝑦 𝑦 = 2 ∑ 𝑊𝑖 𝑦𝑖 𝑦 = ∑ 𝑊𝑖 𝑦

𝑊𝑖 𝑦𝑖

MATERIALS AND METHODS This section describes the materials and procedures used in the design and development of the computer vision-based weed control system for low- land rice precision farming. The method is explained from the prism of system overview, fuzzy logic control, hardware design considerations, software design considerations, and performance evaluation. A computer vision-based weed control system for low-land rice precision farming (O.M. Olaniyi)



ISSN: 2252-8814

3.1. System fuzzy logic control design The fuzzy logic-based weed control system was simulated using Sugeno and Mamdani fuzzy logic models, and was designed using the fuzzy logic designer toolbox of MATLAB. It comprises of two inputs, the weed bounding box size (boxSize) and the confidence level of the recognition model to drive an output which is the spray rate (volume of herbicides sprayed per time). The box size is taken from 0 to 10 and confidence level from 0 to 100. Each of the input has its triangular membership function as shown in Figure 2 and Figure 3 for Sugeno, and Figure 4 and Figure 5 for Mamdani FIS.

Figure 2. Sugeno triangular membership function for box size

Figure 3. Sugeno triangular membership function for confidence level

Figure 4. Mamdani membership function for box size

Figure 5. Mamdani membership function for confidence level

Weighted average was used for obtaining crisp output of Sugeno FIS and centroid method of defuzzification was used for getting out crisp output of Mamdani. The Mamdani FIS has output membership function while the Sugeno does not have, rather it has a table of constant for flexibility. Table 2 and Figure 6 show the output for Sugeno and Mamdani FIS respectively. For this research Sugeno is adopted because it is less time consuming in terms of defuzzification by using weighted average instead of defuzzification as used by Mamdani model. It is more efficient, works well with optimization and has adaptative ability [26]. Table 3 and Table 4 show that both Sugeno and Mamdani uses the same rule set. The rules which set the control are shown in both tables. Table 2. Sugeno output table Spray rate Off Low Medium High

Table 3. System fuzzy control rules for Sugeno

Constant value 0 0.3333 0.6667 1

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 51 – 61

Rules 1 2 3 4 5 6 7 8 9

Box size SMALL SMALL SMALL MEDIUM MEDIUM MEDIUM BIG BIG BIG

Confidence level LOW MODERATE HIGH LOW MODERATE HIGH LOW MODERATE HIGH

Spray rate OFF LOW MEDIUM LOW MEDIUM HIGH LOW MEDIUM HIGH

Int J Adv Appl Sci



ISSN: 2252-8814

Figure 6. Mamdani triangular membership function for spray rate Table 4. System fuzzy control rules for Mamdani Rules 1 2 3 4 5 6 7 8 9

Box size SMALL SMALL SMALL MEDIUM MEDIUM MEDIUM BIG BIG BIG

Confidence level LOW MODERATE HIGH LOW MODERATE HIGH LOW MODERATE HIGH

Spray rate OFF LOW MEDIUM LOW MEDIUM HIGH LOW MEDIUM HIGH

3.2. System hardware design consideration Raspberry pi 3B was used as the microcontroller of the computer vision system. The raspberry Pi 3B is a third-generation single board computer that has processing speed of 1.2GHz enabling it to process the image at fast speed and send the weed image size and confidence level of image recognition to the fuzzy inference system for control. The mechanical unit consists of the five-litre tank, hose and nozzle which are connected to make the spraying tank. Other components used are the DC pump, Raspberry pi 8Mp camera, LEDs and L293N motor driver which are all connected to the raspberry pi for control. The DC pump is powered by 12V DC while the raspberry pi is powered by 5v DC battery. 3.2.1. System mathematical modelling The system model consists of the controller connected in series with the pump. The block diagram of the model system is in Figure 7, while Figure 9 shows the simulink modelling of the system in MATLAB. G(s) represents the pump and in modelled in terms of a DC motor. The input to the model is electrical voltage which in turns gives an output of angular velocity to drive the motor [27]. The angular velocity of the shaft drive determines the rate at which the herbicide comes out of the pump. The model of the DC motor is shown in Figure 8. Table 5 presents the parameter and value.

Figure 7. Block diagram of the system model

Figure 8. Model diagram of the pump (Chen, 2015)

A computer vision-based weed control system for low-land rice precision farming (O.M. Olaniyi)



ISSN: 2252-8814

Where: 𝑖 (𝑡) = Armature current 𝑅 =Armature resistance 𝐿 = Armature inductance 𝑉 (𝑡) = Applied voltage 𝑉 (𝑡) = Back EMF 𝑇 (𝑡) = Motor torque 𝐽 = Rotor moment of inertia 𝐵 = Frictional coefficient From Kirchhoff’s voltage law the equation for the DC motor is 𝑉 (𝑡) = 𝑅 𝑖 (𝑡) + 𝐿

( )

(1)

+ 𝑉 (𝑡)

The back EMF, 𝑉 (𝑡) is proportional to angular velocity. It is given as (2)

𝑉 (𝑡) = 𝑘 𝜔(𝑡)

𝑘 is the constant of the back EMF. The torque of the motor is proportional to armature current, it is given as (3)

𝑇 (𝑡) = 𝑘 𝑖𝑎(𝑡) Where 𝑘 is the constant of the motor torque. The equation of the dc motor can now be written as 𝑉 (𝑡) = 𝑅 𝑖 (𝑡) + 𝐿

( )

(4)

+ 𝑘 𝜔(𝑡)

Using (3), ( )

𝑖𝑎(𝑡) =

(5)

Substituting (3) into (4), we obtain 𝑉 (𝑡) = 𝑅

( )

+ 𝐿

𝑘 𝑉 (𝑡) = 𝑅 𝑇 (𝑡) + 𝐿

( )

+ 𝑘 𝜔(𝑡)

𝑇 (𝑡) + 𝑘 𝑘 𝜔(𝑡)

(6) (7)

The mechanical behaviour of the motor can be described as 𝑇 (𝑡) = 𝐽

( )

(8)

(𝑡) + 𝐵 𝜔(𝑡)

Taking the Laplace transform of (7) and (8) 𝑘 𝑉 (𝑠) = 𝑅 𝑇 (𝑠) + 𝑆𝐿 𝑇 (𝑠) + 𝑘 𝑘 Ω(𝑠)

(9) (10)

𝑇 (𝑠) = 𝑆𝐽 Ω(𝑠) + 𝐵 Ω(𝑠) Putting (10) into (9), we obtain 𝑘 𝑉 (𝑠) = (𝑅 + 𝑆𝐿 ) 𝑆𝐽 Ω(𝑠) + 𝐵 Ω(𝑠) + 𝑘 𝑘 Ω(𝑠) 𝐺 (𝑠) = Ω( ) ( )

(11)

𝑜𝑢𝑡𝑝𝑢𝑡 𝑖𝑛𝑝𝑢𝑡 (

) (

)

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 51 – 61

(12)

Int J Adv Appl Sci



ISSN: 2252-8814

The values from the standard motor equation (Chen, 2015), are shown in Table 7. Table 5. Parameter table Parameter 𝑅 𝐿 𝐽 𝐾 𝐾 𝐵

Value 0.5Ω 0.0015 H 0.00025 N-m/ (rad/s2) 0.05 N-m/A 0.05V/rad/s 0.0001 N-m/ rad/s

Substituting each value into (12), we finally obtain 𝐺 (𝑠) =

(13)

Figure 9. System model simulation in simulink 3.3. System software design consideration The software used in the computer vision system was chosen using criteria such as ease of implementation, how it supports the components selected and how it reacts to flaws in the systems. Python (version 3.6) programming language was used in the design and development of the system because it was the official programming language of Raspberry Pi; highly efficient and fault tolerant in nature. Proteus software (ISIS 8) was used to design the circuit diagram and MATLAB (R2017A) was used for system modelling, simulation and fuzzy logic control because of its modelling and simulation capability and ease of interfacing with other programming languages. 3.4. System performance evaluation Performance evaluation is carried out to know the effectiveness and efficiency of the methods which were adopted in the development of this project. The fuzzy logic model was evaluated using the Mean Absolute Error (MAE) and Mean Square Error (MSE) of the Sugeno model which were the methods as adopted by [28-31]. Database of 50 different samples of box size and confidence level was generated randomly to test the spray rate of the Sugeno and mamdani model using the ‘evalfis’ command in MATLAB. The MAE and MSE were calculated for each deviation of the model. The metrics can be represented mathematically as: MAE=

∑|

(14)

MSE =

∑(

)

(15)

RESULTS AND DISCUSSION This section presents the result obtained from the simulation of both the Sugeno and Mamdani on MATLAB R2017A. Figure 10 shows the surface view for the Sugeno model. Figure 11 shows how the spray rate is affected by the confidence level. The model shows the relationship between the box size of the recognised weed and the confidence level obtained and how they both affect the spray rate of the pump. Figure 14 shows the step response of the designed control system. A computer vision-based weed control system for low-land rice precision farming (O.M. Olaniyi)



ISSN: 2252-8814

Figure 12 and Figure 13 show the surface view for the Mamdani FIS and how the confidence level affects the Spray Rate. It shows how the box size of the recognized weed and the confidence level of the recognition system affects the spray rate of the sprayer. The result obtained for Sugeno model and Mamdani shows that the Sugeno model helps to get the maximum use of the sprayer while the Mamdani shows expressiveness in terms of the output membership function. Figure 15 shows the developed the computer vision system after programming and integrating materials selected in section 3.2.

Figure 10. Surface view for the Sugeno

Figure 11. Spray rate versus confidence level

Figure 12. Surface view for Mamdani

Figure 13. Spray rate vs confidence level

Figure 14. Step response of weed control system

Figure 15. Designed control system

4.1. Performance evaluation for the fuzzy logic model 50 random data were generated to evaluate the Sugeno model and the Mamdani model. The deviation of the model was calculated and recorded. MAE and MSE were the metrics used to evaluate the model. Table 6 and Table 7 show the summary of the results obtained for performance evaluation measurement for 50 generated samples for both Sugeno and Mamdani fuzzy logic system. The equation is adopted from Section 3.4. Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 51 – 61

Int J Adv Appl Sci ∑|

𝑀𝐴𝐸 = MSE =



ISSN: 2252-8814

∑(

)

For Sugeno .

𝑀𝐴𝐸 = 𝑀𝑆𝐸 =

= 0.09135 = 0.034786

For Mamdani 𝑀𝐴𝐸 = 𝑀𝑆𝐸 =

= 0.09054 = 0.02429

Table 6. Summary of performance measurement for the 50 test samples for Sugeno Performance measure MAE MSE

50 Test samples 0.09 0.03

Table 7. Summary of performance measurement for the 50 test samples for Mamdani Performance measure MAE MSE

50 Test samples 0.09 0.02

CONCLUSION The result obtained from the simulation shows 0.09 for mean absolute error and 0.03 for mean square error of Sugeno, similar value was obtained for the evaluation of Mamdani fuzzy inference system. This result shows that fuzzy logic can be applied to monitor the spraying action of a tank which in turn will lead to reduction in herbicide usage and reduce environmental damage associated with uniform spraying. The system modelling on Simulink shows a rise time of 0.1012, settling time of 0.1829, overshoot of 0.0045 and peak time of 2.3734. Since the system error is negligible, it shows that the system can be applied on a real rice farm to control herbicide usage, thus boosting rice production at reduced production cost. This will therefore yield more returns for the farmers. However, in this research, the actual amount of herbicide to kill the weeds was not calculated. Therefore, future research should endeavour to carry out quantitative amount of herbicides capable of killing the recognized weeds. Furthermore, the sprayer arm for the system was static, so future work can work a robotic hand for the sprayer to enable it turn and spray in respect to viewed areas. REFERENCES [1]

R.K. Naresh, Y. Kumar, P. Chauhan, and D. Kumar, “Role of precision farming for sustainability of rice-wheat cropping system in western indo gangetic plains,” Int J Life Sc Bt Pharm Res, vol. 1, pp. 1–13, 2012. [2] Shibusawa S., “Precision Farming and Terra-Mechanics,” The Fifth ISTVS Asia-Pacific Regional Conference in Korea, October 20-22, pp. 251-261, 1998. [3] Amin, M. S. M, Rowshon, M. K., and W. Aimrun, “Paddy water management for precision farming of rice,” Current Issues of Water Management, InTech, 2011. [4] Maclean, J., Hardy, B., and Hettel, G., Rice Almanac: Source book for one of the most important economic activities on earth. IRRI, 2013. [5] Nwilene, F., Oikeh, S., Agunbiade, T., Oladimeji, O., Ajayi, O., Sié, M., and Touré, A., Growing lowland rice: a production handbook. Africa Rice Center (WARDA), 2008. [6] Rodenburg, J. and Johnson, D. E., “Weed Management in Rice‐Based Cropping Systems in Africa,” Advances in Agronomy, D. L. Sparks, Ed. Academic Press, vol. 103, pp. 149-218, 2009. [7] Khan, Y. N., “Weed Detection in Crops Using Computer Vision,” Centre of Robotics, 2015. [8] Hossain, M. M., “Recent perspective of herbicide: Review of demand and adoption in world agriculture,” Journal of Bangladesh Agricultural university, vol. 13, no. 1, pp. 19–30, 2015. [9] Ismaila, U., Wada, A. C., Daniya, E., and Gbanguba, A. U., “Meeting the Local Rice Needs in Nigeria Through Effective Weed Management,” Sustainable Agriculture Research, vol. 2, no. 2, pp. 37–48, 2013. [10] Patel, D. D., and Kumbhar, B. A., “Weed and its management: A major threats to crop economy,” J. Pharm. Sci. Bioscientific Res, vol. 6, no. 6, pp. 453–758, 2016.

A computer vision-based weed control system for low-land rice precision farming (O.M. Olaniyi)



ISSN: 2252-8814

[11] Ajala, A. S., and Gana, A., “Analysis of Challenges Facing Rice Processing in Nigeria,” Journal of Food Processing, vol. 2015, pp. 1-6, 2015 [12] Agro Nigeria, “Weed Control: All You Need to Know About Using Simazine,” 2014. Retrieved May 30, 2018. [Online]. Available: from https://agronigeria.com.ng/weed-control-need-know-using-simazine/ [13] Bajwa, A. A., Mahajan, G., and Chauhan, B. S., “Nonconventional weed management strategies for modern agriculture,” Weed Science, vol. 63, no. 4, pp. 723–747, 2015. [14] Patel, D. D., and Kumbhar, B. A., “Weed and its management: A major threats to crop economy,” J. Pharm. Sci. Bioscientific Res, vol. 6, no. 6, pp. 453–758, 2016. [15] Vikhram, G. Y. R., Agarwal, R., Uprety, R., and Prasanth, V. N. S., “Automatic Weed Detection and Smart Herbicide Sprayer Robot,” International Journal of Engineering and Technology (UAE), vol. 7, pp. 115-118, 2018. [16] Hassan, N. M., “Development of Intelligent Fruit Pulp Recognition System Using Statistical Descriptor and Artificial Neural Network,” Master of Engineering Thesis, Department of Computer Engineering. Federal University of Technology, Minna, Nigeria, 2015. [17] Singh, V., Jat, M. L., Ganie, Z. A., Chauhan, B. S., and Gupta, R. K., “Herbicide options for effective weed management in dry direct- seeded rice under scented rice-wheat rotation of western Indo- Gangetic Plains,” Crop Protection, vol. 81, pp. 168–176, 2016. [18] Deepa, K., and Sujatha, N., “Analysis and Detection of Weeds in Agricultural Area using various Image Segmentation Algorithms,” International Journal of Scientific Engineering and Research (IJSER), vol. 5, no. 7, pp. 256–260, 2017. [19] Desai, R., Desai, K., and Solanki, Z., “Removal of Weeds Using Image Processing: A Technical Review,” International Journal of Advanced Computer Technology (IJACT), vol. 4, no. 1, pp. 27-31, 2015. [20] Pusphavalli, M., and Chandraleka, R., “Automatic Weed Removal System using Machine Vision,” International Journal of advanced Research in Electronics and Communication Engineering, vol. 5, no. 3, pp. 503-506, 2016. [21] Pulido, C., Solaque, L., and Velasco, N., “Weed recognition by SVM texture feature classification in outdoor vegetable crop images,” Ingeniería e Investigación, vol. 37, no. 1, pp. 68–74, 2017. [22] Tutorial Point., “Artificial Intelligence - Fuzzy Logic Systems,” 2018. Retrieved September 12, 2018. [Online]. Available: https://tutorialpoints.com/artificial_intelligence/artificial_intelligence_fuzzy_logic_systems.htm [23] Wang, Y., and Chen, Y., “A Comparison of Mamdani and Sugeno Fuzzy Inference Systems for Traffic Flow Prediction,” Journal of Computers, vol. 9, no. 1, pp. 12–21, 2014. [24] Sari, W. E., Wahyunggoro, O., and Fauziati, S., “A Comparative Study on Fuzzy Mamdani-Sugeno- Tsukamoto for the Childhood Tuberculosis Diagnosis,” IP Conf Proc., vol. 1755, 2016. [25] Zaher, H., Kandil, A. E., and Fahmy, R., “Comparison of Mamdani and Sugeno Fuzzy Inference Systems for Prediction,” Science Domain International, vol. 4, no. 21, pp. 3014–3022, 2014. [26] Kaur, A., and Kaur, A., “Comparison of Mamdani-Type and Sugeno-Type Fuzzy Inference Systems for Air Conditioning System,” International Journal of Soft Computing and Engineering (IJSCE), vol. 2, no. 2, pp. 323–325, 2012. [27] Y. Chen, “Modeling of DC Motor,” pp. 1–8, 2015. [28] Chadha, A., and Satam, N., “An Efficient Method for Image and Audio Steganography Using Least Significant Bit (LSB) Subtitution.,” International Journal of Computer Applications, vol. 77, no. 13, pp. 37-45, 2013. [29] Isah, A.D., “Development of Asphalt Paved Road Pothole Detection System Using Modified Colour Space Approach,” M.Eng. Thesis, Department of Computer Engineering. Federal University of Technology, Minna, Nigeria, 2015. [30] Gupta, H., Kumar, R., and Changlani, S., “Enhanced Data Hiding Capacity Using LSB- Based Image Steganography Method,” International Journal of Emerging Technology and Advanced Engineering, vol. 3, no. 6, pp. 212-214, 2013. [31] Laskar, S. A. and Hemachandran, K., “High capacity data hiding using LSB steganography and encryption,” International Journal of Database Management Systems (IJDMS), vol. 4, no. 6, pp. 57-68, 2012.

BIOGRAPHIES OF AUTHORS Olayemi Mikail Olaniyi is an Associate Professor in the Department of Computer Engineering at Federal University of Technology, Minna, Niger State, Nigeria. He obtained his B.Tech. and M.Sc. in Computer Engineering and Electronic and Computer Engineering respectively. He had his Ph.D. in Computer Security from Ladoke Akintola University of Technology, Ogbomosho, Oyo State, Nigeria. He has published in reputable journals and learned conferences. His areas of research include Computer Security, Intelligent/Embedded Systems design and Applied Medical Informatics

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 51 – 61

Int J Adv Appl Sci

ISSN: 2252-8814



Emmanuel Daniya received a B Agric. Tech., Crop Production, M.Tech. (Agronomy) degrees from the Federal University of Technology, Minna in 1998 and 2004 respectively; and completed his Ph.D. (Agronomy) degree in 2014 from Ahmadu Bello University, Zaria, Nigeria; with specialization in Weed Science. His area of interest include Weed Science, Cereal and Legume Crops Production, Farming Systems, and Statistical Methods, Experimental Design, weed biology, ecology and weed management strategies. He has his research findings published in reputable national and international peer reviewed journals Engr. Dr. Jonathan Gana Kolo is an Associate Professor in the Department of Electrical and Electronics Engineering.He received Bachelor of Engineering (B.Eng.) Degree in Electrical Engineering from Ahmadu Bello University, Zaria; Master of Science (M.Sc.) Degree in Electrical and Electronic Engineering from the University of Lagos, Lagos; Doctor of Philosophy (PhD) Degree from The University of Nottingham Malaysia Campus. His research interests are majorly in the areas of Wireless Sensor Networks, Embedded Systems, Electronics, Digital Signal Processing, Intelligent Systems, Digital Image processing and Communication Engineering. His Research results are published in several peer- reviewed papers. Jibril Abdullahi Bala is a Graduate Assistant with the Department of Mechatronics Engineering, Federal University of Technology, Minna, Nigeria. He obtained his Bachelor’s degree in Computer Engineering from Federal University of Technology, Minna, Nigeria. His area of interests are Control, Artificial Intelligence and Embedded Systems.

Ayobami Esther Olanrewaju is a Junior consultant on SAP/ABAP practices at Thamani MultiConcept, Lagos, Nigeria. She obtained her Bachelor’s degree in Computer Engineering from Federal University of Technology, Minna, Nigeria. Her area of interests are Data Science, Artificial Intelligence, Big data and Analytics, and Machine learning.

A computer vision-based weed control system for low-land rice precision farming (O.M. Olaniyi)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 62~66 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp62-66



Intrusions detection using optimized support vector machine Mehdi Moukhafi, Khalid El Yassini, Bri Seddik Department of Mathematics and Computer Sciences, Faculty of Sciences, Moulay Ismail University, Morocco

Article Info

ABSTRACT

Article history:

Computer network technologies are evolving fast and the development of internet technology is more quickly, people more aware of the importance of the network security. Network security is main issue of computing because the number attacks are continuously increasing. For these reasons, intrusion detection systems (IDSs) have emerged as a group of methods that combats the unauthorized use of a network’s resources. Recent advances in information technology, specially in data mining, have produced a wide variety of machine learning methods, which can be integrated into an IDS. This study proposes a new method of intrusion detection that uses support vector machine optimizing optimizing by a genetic algorithm. to improve the efficiency of detecting known and unknown attacks, we used a Particle Swarm Optimization algorithm to select the most influential features for learning the classification model.

Received May 3, 2019 Revised Oct 20, 2019 Accepted Jan 15, 2020 Keywords: Genetic algorithm Intrusion detection system kdd99 Particle swarm optimization Support vector machine

This is an open access article under the CC BY-SA license.

Corresponding Author: Mehdi Moukhafi, Informatics and Applications Laboratory (IA), Department of Mathematics and Computer Sciences, Faculty of Sciences, Moulay Ismail University, Meknes, Morocco. Email: mehdi.moukhafi@gmail.com

INTRODUCTION Due to the tremendous growth of the Internet and Network based services, the severity o f network based computer attacks have significantly increased. Thus, an intrusion detection system (IDS) play a vital role in network security. Intrusion detection system tries to detect computer attacks by examining various data records. the IDS was presented for the first time by Anderson in 1980 [1], and later formalized by Denning [2], can be used in the global security politics, which includes other protection tools, such as firewalls and anti-virus software; Thus, it is important to take the advantage of these tools collaboration and complementarity. Actual IDS’s based on heuristic rules, such as Snort are signature based system. The problem with such a system is that it cannot detect novel attacks whose signature is not available and hence generates a high rate of alerts. where constantly the environments were changing, the major drawback of approaches based signature is that they only detect known attacks, which implies a frequent updating of the rules database and the time for the implement. To overcome the mentioned problem above, many data mining techniques have been developed [3]. The data mining techniques are better applied equally for an anomaly intrusions detection, also for a knowledge-based intrusions detection [4]. The statistical analysis of the normal system behavior is one of the first approaches to intrusion detection. The statistics are used mathematically to describe an observed mechanism. Generally, the observations allow to get a rough description. For this, the value of certain observations is considered random variables. For each of its comments, a statistical model is used to describe the set of the corresponding random variable distributions. Journal homepage: http://ijaas.iaescore.com

Int J Adv Appl Sci

ISSN: 2252-8814



Learning algorithms can play an important role in detecting attacks (known or unknown). Additionally, the IDS’s performances is considerably improved at the network level. SVM obtains a good detection performance in terms of classifying the flow of a network into normal or abnormal behaviors. Feng et al. [5] introduced an approach combining SVM with self-organized ant colony network. Kuang [6] propose a solution based on a combination of the SVM model with kernel principal component analysis (KPCA) and genetic algorithm. KPCA was used to reduce the dimensions of feature vectors, whereas GA was employed to optimize the SVM parameters. Al-Yaseen et al. [7] Propose a solution based on hybrid SVM and Extreme Learning Machine model Learned with data set built by a modified K- means The modified K-means is used to build new small training datasets representing the entire original training dataset. The rest of this work is organized as follows: Section 1 describes the PSO, SVM and GA methods, section 2 proposed architecture, section 3 simulation of results and evaluation of the algorithm. Section 4 conclusion and future work. 2. THE USED METHODS 2.1. Particle swarm optimization Particle Swarm Optimization (PSO) is a stochastic optimization method, for the nonlinear functions, inspired by the social behavior of insect colonies, bird flocks, fish schools and other animal societies, PSO was invented by Russell Eberhart and James Kennedy [8] in 1995. Originally, the two began developing software simulations birds flocking around food sources, later after realizing that their algorithm solve optimization problems, they present [9] a discrete binary PSO algorithm developed from the previous PSO and operating in continuous variables. PSO is an iterative algorithm to find the best solution based on a population composed of many particles. For example, a flock of birds (particles) encircling an area where they can feel a hidden source of food. Whoever the closest to food warn others birds to move toward its direction. If any of the other birds circling closer to the target more than the first, it warbles stronger and the others move towards him. This scheme continues until one of the birds find food. A particle (candidate solution) that may move to the optimal position by updating its position and its speed. The speed of movement of a particle can be updated by the weight of inertia, cognitive learning factor, and the values of social learning factors. 2.2. Support vector machine Support Vector Machine (SVM) is one of the most popular supervised machine learning algorithms. This is a classification model by evaluating data and identify patterns that retains excellent long generalization capabilities with an integrated resistance to overtraining. This generalization is based on solid theoretical foundations introduced by Vapnik [10]. An SVM model as shown in Figure 1 is an illustration of examples of points in two-dimensional space, where instances of different groups are separated by an area called margin.

Figure 1. Classical example of SVM linear classifier In the classification of support vector, the separation function is a linear combination of grains as given in (1) and are in contact with the support vector f(x) = ∑ ∈ μ y x x + b

(1)

where μ is a Lagrange xi factor is training models, yi {+ 1, -1} is the corresponding class labels and S denotes the set of support vectors. Intrusions detection using optimized support vector machine (Mehdi Moukhafi)



ISSN: 2252-8814

2.3. Genetic algorithm Genetic Algorithm (GA) [11] a general adaptive optimization search methodology based on a direct analogy to Darwinian’s principle of evolution and survival of fittest to optimize a population of candidate solutions towards a predefined fitness. The procedure for a genetic algorithm is:  Initialization: Create an initial population. This population is usually randomly generated of n chromosomes (suitable solutions for the problem)  Evaluation: Each chromosome x of the population is then evaluated and we calculate a 'fitness' for that individual.  Selection: Select two parent chromosomes from a population, based on their fitness (the better fitness, the bigger chance to be selected)  Crossover: create new individuals by combining aspects of our selected parent.  Mutation: The algorithm creates mutation children by randomly changing the genes of individual parents. Mutation typically works by making very small changes at random to an individual genome.  Test: If the end condition is satisfied, stop, and return the best solution in current population, else return to selection step. 3. PROPOSED MODEL FOR INTRUSION DETECTION 3.1. PSO features selection From an artificial intelligence perspective, create a classifier means creating a template for data, or perfect for a model is to be as simple as possible. Reducing the number of parameters, then reduces the number of parameters necessary for the description of this model.  It improves the classification performance: learning time, his speed and power of generalization.  It increases the comprehensibility of data. This data selection is to select an optimum subset of relevant variables from a set of original variables. The KDD99 has 41 variables, it is relatively a large number to be processed by the classifier, the latter in the learning phase cannot complete execution within a reasonable time, then the selection can reduce the feature space. We used PSO to select the optimum features, that have the most impact on the prediction of the model, so we reduce the number of fields from 41 to 16 Table 1. Selected features: 2, 3, 4, 5, 6, 8, 11, 14, 23, 26, 29, 30, 35, 36, 37, 38 Table 1. Comparison of number of features between the data set and subset Data set

Subset

0, 1, 18, 10, 491, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 0, 0, 0, 0, 1, 0, 0, 150, 25, 0.17, 0.03, 0.17, 0, 0, 0, 0.05, 0, normal

1, 18, 10, 491, 0, 0, 0, 0, 2, 0, 1, 0, 0.03, 0.17, 0, 0, normal

3.2. Architecture of proposed IDS To implement our proposed approach, the RBF kernel function is used for the SVM classifier because the RBF kernel function has an excellent performance for the management of higher-dimensional data and requires that only two parameters:  C (penalty parameter): This parameter, common to all SVM kernels, trades off misclassification of training examples against simplicity of the decision surface.  (gamma parameter): It is a specific parameter to RBF kernel function, gamma defines how much influence a single training example has. The parameters (C and g) used as input attributes must be optimized using genetic algorithm. To precisely establish a GA-SVM based intrusion detection system, the following main steps (as shown in Figure 2) must be proceeded. The detailed explanation is as follows: Step 1: Features selection using PSO for a training data set (kdd99_p) Step 2: Initialization of population by a genetic algorithm Step 3: Genetic algorithm process: In this step, the system searches for better solutions by genetic operations, including selection, crossover, mutation (Described in the previous session) Step 4: Training SVM classifier using optimized parameters (C,) Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 62 – 66

Int J Adv Appl Sci



ISSN: 2252-8814

Step 5: Fitnessevaluation. For each chromosome representing C, , training dataset is used to train the SVM classifier, each chromosome is evaluated by fitness function Step 6: Termination criteria :When the termination criteria are satisfied, we evaluate a model using a full kdd99 data set; otherwise, return to step 3.

Figure 2. Proposed algorithm 4.

EXPERIMENT SETUP AND PERFORMANCE EVALUATION In this section, we evaluate the performance of the proposed model. All experiments were conducted on a calculation station 24 CPU Intel Core 2.13GHz, 48GB RAM, running under Linux CentOS 7. The implementation was coded using the Java language. 4.1. Data set Cyber Systems and Technology Group of MIT Lincoln Laboratory [12] simulated LAN US Air Force LAN with multiple attacks and captured nine weeks TCPdump data. This database was first used for competitions kdd99, but since it has become the database test to the IDS’s based on a behavioral approach. KDD Cup 1999 provided both the training dataset, it is called KDd99_10p. Each connection record consists of approximately 100 bytes. This was converted into about 49 * 105 connection vectors each one contains 41 fields. This database is collected by simulating attacks on different platforms such as Windows, Unix, etc ... Four gigabytes of raw data compressed TCP dump is transformed into five million connections files. The attacks are divided into four main categories: Denial of Service Attack (DOS), Probing, User to Root Attack (U2R), Remote to Local Attack (R2L). 4.2. Anomaly detection results This section describes the obtained results from the experiment by applying the proposed algorithms on the data set kdd99. The performance of the proposed method of intrusion detection was evaluated on all KDD99 data set, 10% of the KDD99 data set were used for training the GA-SVM model after the features selection by PSO. Tables 2 illustrates the confusion matrix. this system achieves a top performance of up to 96,01% with a reasonable false alarm rate of 0,02% and a detection rate of 96,38%. Table 2. Confusion matrix Actual Class Normal DOS Probe R2L U2R

Normal 713680 580 519 4 0

Classified Class DOS Probe R2L 106761 26767 30811 3878986 3784 6 6646 33797 3 0 17 923 0 0 21

U2R 15762 14 137 182 31

Intrusions detection using optimized support vector machine (Mehdi Moukhafi)



ISSN: 2252-8814

Figure 3 shows the detection rate classified by attack, the proposed algorithm has detected 99,89% of DOS attacks whom are the most used by hackers. For Probe attacks a rate of 82,23% is correctly classified, 81,97% for R2L attacks and 59,62% for U2R, the low rate of detection U2R attacks can be explained by the insufficient number of data record-learning. To compare a detection rate, Figure 4 compares the proposed method with approaches that use only the entire KDD Cup 1999 dataset as a testing dataset because several researchers used only part of the KDD Cup 1999. The above results show that our approach enhances the performance of IDS. GA-SVM with PSO selection features is more reliable than state- ofthe-art methods. The strengths of the proposed method are the highly improved detection accuracy compared with other methods because of the high reduction of original training dataset size which simplified the learning phase of the classifier and the SVM parameters optimisation by GA.

Figure 3. The accuracy rates per attack

Figure 4. Comparison of proposed model with other methods by detection rate

CONCLUSION In this paper, a Novel hybrid GA-SVM with PSO feature selection–based is proposed for intrusion detection. The proposed model is marked by a significantly better performance. RBF kernel function is used for improve the performance of SVM classification model, GA is used to select suitable parameters for SVM classifier and PSO is designated to select a features of the training dataset and provide new high-quality training datasets that can improve the overall performance of GA-SVM. We also carried out comparisons of our method against other methods, and have shown a noticeable performance. For future work, we want to develop more approaches to combine several machine learning techniques into one predictive model, using meta-algorithms, to increase the rate of detection of attacks. REFERENCES [1]

J.P. Anderson, Computer security threat monitoring and surveillance. Technical Report, Fort Washington, PA, USA, 1980. [2] D. E. Denning, “An Intrusion-Detection Model,” IEEE Transactions on Software Engineering, Vol. 13, No. 2, pp. 222–232, 1987. [3] S. Forrest, S.A. Hofmeyr, A. Somayaji, and T.A. Longstaff, “A sense of self for Unix processes,” Proceedings of IEEE Symposium on Security and Privacy, Washington, pp. 120-128, 1996. [4] W.L.W. Lee, SJ, and Mok KW., “A data mining framework for building intrusion detection models,” Proceedings of the 1999 IEEE Symposium on Security and Privacy, California, pp. 120–132, 1999. [5] W. Feng, Q. Zhang, G. Hu, and J.X. Huang, “Mining network data for intrusion detection through combining SVMs with ant colony networks,” Future Generation Computer Systems, Vol. 37, pp. 127–140, 2014. [6] F. Kuang, W. Xu, and S. Zhang, “A novel hybrid KPCA and SVM with GA model for intrusion detection,” Applied Soft Computing Journal, Vol. 18, pp. 178–184, 2014. [7] W.L. Al-Yaseen, Z.A. Othman, and M.Z.A. Nazri, “Multi-level hybrid support vector machine and extreme learning machine based on modified K-means for intrusion detection system,” Expert Systems with Applications, Vol. 67, pp. 296–303, 2017. [8] J. Kennedy and R. Eberhart, “Particle swarm optimization,” Proceedings of IEEE International Conference,Vol. 4, pp. 1942–1948, 1995. [9] J. Kennedy and R. Mendes, “Population structure and particle swarm performance,” Proceedings of the 2002 Congress on Evolutionary Computation, Vol. 2, pp. 1671–1676, 2002. [10] C. Cortes and V. Vapnik, “Support Vector Networks,” Machine Learning, Vol. 20, No. 3, pp. 273–297, 1995. [11] O. Kramer, Genetic Algorithm Essentials. Cham: Springer International Publishing, 2017. [12] M. Tavallaee, E. Bagheri, W. Lu, and A.A. Ghorbani, “A detailed analysis of the KDD CUP 99 data set,” in EEE Symposium on Computational Intelligence for Security and Defense Applications, pp. 1–6, 2009.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 62 – 66

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 67~69 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp67-69



Top-K search scheme on encrypted data in cloud K. Pushpa Rani1, L. Lakshmi2, Ch. Sabitha3, B. Dhana Lakshmi4, S. Sreeja5 1,2Department

of Computer Science and Engineering, MLR Institute of Technology, India Department of Computer Science and Engineering, Vardhaman College of Engineering, India 4 Department of Computer Science and Engineering, IARE College of Engineering, India 5 Department of Computer Science and Engineering, St. Martins College of Engineering, India

Article Info

ABSTRACT

Article history:

A Secure and Effective Multi-keyword Ranked Search Scheme on Encrypted Cloud Data. Cloud computing is providing people a very good knowledge on all the popular and relevant domains which they need in their daily life. For this, all the people who act as Data Owners must possess some knowledge on Cloud should be provided with more information so that it will help them to make the cloud maintenance and administration easy. And most important concern these days is privacy. Some sensitive data exposed in the cloud these days have security issues. So, sensitive information ought to be encrypted earlier before making the data externalized for confidentiality, which makes some keyword-based information retrieval methods outdated. But this has some other problems like the usage of this information becomes difficult and also all the ancient algorithms developed for performing search on these data are not so efficient now because of the encryption done to help data from breaches. In this project, we try to investigate the multi- keyword top-k search problem for encryption against privacy breaks and to establish an economical and secure resolution to the present drawback. we have a tendency to construct a special tree-based index structure and style a random traversal formula, which makes even identical question to supply totally different visiting ways on the index, and may additionally maintain the accuracy of queries unchanged below stronger privacy. For this purpose, we take the help of vector area models and TFIDF. The KNN set of rules are used to develop this approach.

Received May 8, 2019 Revised Sep 20, 2019 Accepted Dec 10, 2019 Keywords: Cloud Data proprietors Multi keyword ranked seek over encrypted cloud facts OTP Product resemblance

This is an open access article under the CC BY-SA license.

Corresponding Author: K. Pushpa Rani, Department of Computer Science and Engineering, MLR Institute of Technology, Hyderabad-500043, India. Email: rani536@gmail.com

INTRODUCTION These days, cloud computing [1] has emerged as an essential mechanism for plenty utilities, where cloud customers can keep their statistics into the cloud that allows them to take advantage from on-demand extremely good request and offerings from a shared pool of configurable computing assets. Cloud computing is now a days a trend in most of the IT industries because of its extra ordinary features like Pay-as-you-go basis. This will help people in achieving their necessities with basic cost on all the resources. People need not spend so much in the starting stage itself. As in the beginning, any company needs only basic services and based on its growing demand, it can set up all the resources and necessities. Hence cloud computing has become a good trend as it is eliminating most of the unnecessary costs to a basic startup company. Nowadays, additional and additional corporations and people from an outsized variety of huge information Journal homepage: http://ijaas.iaescore.com



ISSN: 2252-8814

applications have source their information and maintain their data in cloud servers for simple information management, economical process and question processing tasks. In such cases there is a high risk of security [2, 3] issues as there are many sensitive data like e-mail, health records of individuals etc. The owner of the data [4] is concerned about privacy apart from enjoying the benefits of the cloud. Their outsourced information must be in a way that is more secure so that it is not possible for illegitimate users [5-8] to access their data. For this purpose, the outsourced data must be converted to a format that is not readable easily but is able to accessible only to those who is a legal or valid user of the data. In our project, this can be done by encrypting the data before it is put in the cloud. And as a result of this, the ancient methods for data retrieval are not so efficient on such data. So, there is another method proposed in the project that makes the Search on this type of data fast and efficient. It is known as Top-k approach [9, 10]. This approach will help construct tree-based indexes which are nearer to search criteria. Different tree-based paths are obtained which when traversed gives unique search results. 2.

RESEARCH METHOD The framework consists of explorable encoded method that helps the accurate multiple key word ranked seek and bendy vigorous running on document group. This framework is a relaxed tree shaped based totally exploring model on the enciphered cloud statistics, which helps multi key word ranked seek. The so-called vector space model and the extensively casted off “time period frequency (TF) × inverse record frequency (IDF)” groups are binded to the index construction and also for generation of question of search. Algorithm: Term Frequency-Inverse Document Frequency Input: Data d. Output: result r. Let data d, Collection c; c=getWords(d); //Using Split(“\\s+”) Term Frequency tf1; α= the count of terms t appearing in a document; tf1=( α);. Inverse Document Frequency idf1; α= The count of the terms t that are present in a document; β= Total number of terms in the document; IDF1(t) = ( α)/( β);. End; 3.

RESULTS AND ANALYSIS The suggested one, data users/people can acquire specific necessities on search correctness of privateness with the aid of the standard deviation of adjustment that can be dealt with as a compensation parameter. The assessment of structures with a recent painting prove that it gives a high seek performance. PMRSE scheme calls the hunt results with the aid of specific reckoning of two types of vectors i.e document and query. Thus, the seek accuracy of PMRSE scheme is 100%. But based totally and similarity Multi- keyword rectangular seek pattern, the basic scheme is affected by lack of precision because of various factors like accumulation of sub-vectors along with the index creation. The validation is iterated 16 number of times. Average accuracy of 91%. During the quest, whilst the relevance of the node is higher in Rlist, examines the server of the cloud. Because it is a balanced binary tree, its height of the index n should be taken care. The convolution of the calculation is ranked relevance order of m. We carried out an experimental assessment of the existing System of RSSE and the proposed one PMRSE. The Comparison graph is drawn. The graph coordinates are based on number of documents the corresponding system’s search end result given back along with the time required to go back the documents. The complete experiment machine is carried out with the aid of Java on a Windows Server with Intel i5 2.93GHz. Figure 1 – axis will show the Time (milliseconds), Y- axis represents no. of documents retrieved. By this comparision PMRSE got best results compare with RSSE. The result values is presented in Table 1.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 67 – 69

Int J Adv Appl Sci



ISSN: 2252-8814

Figure 1. Comparison PMRSE with RSSE Table1. Results table Time (ms) 100 250 500 750 1000 1250 1500

RSSE 55 77 103 88 111 120 128

PMRSE 78 109 134 124 144 167 200

CONCLUSION AND FUTURE WORK The proposed model tries to improve the coherence of the top- k multiple keyword search over encrypted data. For this purpose, we try two same questions with unalike keys, for which the server traverse through two unassociated byways to give the user most accurate search results. Then, we also tried to divide the entire dictionary into multiple groups top-ck documents while building index. Traversal algorithm used is RGTMS. Finally, the experimental upshots will teach that our methods are extra added efficient along with a safer than the state-of-the-art methods. REFERENCES [1]

K. Ren, C. Wang, Q. Wang, et al., “Security challenges for the public cloud,” IEEE Internet Computing, vol. 16, no. 1, pp. 69–73, 2012. [2] S. Kamara and K. Lauter, “Cryptographic cloud storage,” in Financial Cryptography and Data Security, Springer, pp. 136–149, 2010. [3] C. Gentry, “A fully homomorphic encryption scheme,” Ph.D. dissertation, Stanford University, 2009. [4] O. Goldreich and R. Ostrovsky, “Software protection and simulation on oblivious rams,” Journal of the ACM (JACM), vol. 43, no. 3, pp. 431–473, 1996. [5] D. Boneh, G. Di Crescenzo, R. Ostrovsky, and G. Persiano, “Public key encryption with keyword search,” in Advances in Cryptology-Eurocrypt 2004, Springer, pp. 506–522, 2004. [6] D. Boneh, E. Kushilevitz, R. Ostrovsky, and W. E. Skeith III, “Public key encryption that allows pir queries,” in Advances in Cryptology-CRYPTO 2007. Springer, pp. 50–67, 2007. [7] D. X. Song, D. Wagner, and A. Perrig, “Practical techniques for searches on encrypted data,” in Security and Privacy, 2000. S&P 2000.Proceedings, 2000 IEEE Symposium, pp. 44–55, 2000. [8] E.-J. Goh et al., “Secure indexes,” IACR Cryptology ePrint Archive, vol. 2003, p. 216, 2003. [9] Y.-C. Chang and M. Mitzenmacher, “Privacy preserving keyword searches on remote encrypted data,” in Proceedings of the Third international conference on Applied Cryptography and Network Security. SpringerVerlag, pp. 442–455, 2005. [10] R. Curtmola, J. Garay, S. Kamara, and R. Ostrovsky, “Searchable symmetric encryption: improved definitions and efficient constructions,” in Proceedings of the 13th ACM conference on Computer and communications security. ACM, pp. 79–88, 2006.

Top-K search scheme on encrypted data in cloud (K. Pushpa Rani)

International Journal of Advances in Applied Sciences (IJAAS) Vol. 9, No. 1, March 2020, pp. 70~76 ISSN: 2252-8814, DOI: 10.11591/ijaas.v9.i1.pp70-76



Software defined network emulation with OpenFlow protocol Tsehay Admassu Assegie Department of Computing Technology, Aksum University, Ethiopia

Article Info

ABSTRACT

Article history:

In software defined network the network infrastructure layer where the entire network devices, like switches and routers reside is connected with the separate controller layer with the help of standard called OpenFlow. The open flow standard enables different vendor devices like juniper, cisco and Huawei switch to connect to the controller or a software program. The software program controls and manages the network devices. Therefore, software defined network architecture makes the network flexible, cost effective and manageable, enables dynamic provisioning of bandwidth, dynamic scale out and dynamic scale in compared to the traditional network. In this study, the architectures and principles of software defined network is explored by emulating the software defined network employing a mininet.

Received Nov 9, 2019 Revised Jan 10, 2020 Accepted Feb 3, 2020 Keywords: Mininet OpenFlow POX controller SDN emulation Software defined network

This is an open access article under the CC BY-SA license.

Corresponding Author: Tsehay Admassu Assegie, Department of Computing Technology, College of Engineering and Technology, Aksum University, 1010 Aksum Universit, Aksum, Ethiopia. Email: tsehayadmassu2006@gmail.com

INTRODUCTION The number of devices connecting to the network increases day by day but, capability of the routing table is limited, a traditional internet protocol (IP) network become increasingly difficult to manage the devices’ in the network and configuration errors have become common problems in addition to the network management problems. The administrator has to go to every device in the network and issue usually a vendor specific commands to set policies, routing information and many network parameters required for the networking device to function and operate smoothly. The complexity of device management had therefore brought the researchers to divert their attention to a new networking paradigm, the software defined network principle. In the traditional network, the software is bundled with the hardware, this bundled technique is called integrated approach and this is costly and even takes time to converge if any failure occurs in the network. The interfaces and the command used in device management are vendor-specific. The bundling of software with the hardware has made it difficult to manage the device and as a solution to these problems a new approach has emerged which is called software defined network. In this approach, the software which is used in device functionality management is separated from the hardware. Software defined network is a networking architecture where the control logic and forwarding functionalities are separated into different layers, the controllable program and the physical infrastructure layers. The key principles of software defined network are: Data plane- is the layer responsible for end to end delivery of data in the network, control plane-deals with IP routing and forwarding decisions, programmable network centrally managed- the management functionalities are centralized, open interfaces between the device in the control plane and the data plane. Journal homepage: http://ijaas.iaescore.com

Int J Adv Appl Sci

ISSN: 2252-8814



RELATED WORKS In this section, the research works carried by different researchers related to a software defined network based on the open flow standard, the architectures of software defined networks, the principles of software defined networks and practical implementation issues of software defined network are reviewed [1-25]. Many researches have been carried out by different scholars on the implementation, challenges and future directions but still the software defined network is in its infancy and further researches are required on this emerging virtualized network infrastructure. A study on emulation of software defined networks indicated that mininet is a powerful software defined network emulation tool [1-10]. Although it is been shown in the study that the mininet is a best tool to emulate software defined networks, the authors used a limited number of hosts in their experiment. This shows that another study to be carried out to test how the emulator behaves in a large-scale network. The software defined network is an emerging trend in the field of networking. With this new emerging trend, the traditional integrated network is assumed to be decoupled into the control plane where the management and functionalities of devices in the data plane, like forwarding packets, policy making in the network is controlled at this layer. And the routing and data forwarding plane is separated [2-11] form the control plane. A software defined network is programmable network where the control logic, which is responsible for the configuration and device management in traditional network, is centralized [12-15] into separate plane called control plane. In traditional networks, the administrator or network operators have to move to each device location and configure the devices according to predefined policy of software integrated into the device. In software defined networks all the control logic is separated from the networking devices like switches and routers and is placed in a layered called the control panel. In software defined network, the open flow is the standard used to forward packets towards their destinations [14-20]. The open flow switches make forwarding decision based on the flow tables. The open flow is the network abstraction layer which defines the standard protocol for communication in the network. It allows the network infrastructure layer to be connected with the control layer. This acts as a bridge between the networking devices and the software defined network controller program. The migration of control logic, which is used to be strongly integrated in the networking devices (for example, Ethernet switches and routers) into accessible and logically centralized controllers, enables the underlying networking infrastructure to be abstracted from the application's point of view. This separation provides a more flexible, programmable, vendor-agnostic, cost effective and innovative network architecture. Software defined network platform provides an effective solution to cost-effective high-speed network services when compared with the traditional network [15-25]. The pox controller is used to manage the device in network infrastructure and they are connected with the help of standard protocol called the open flow. Devices for different vendors may be used in a software defined network using this open flow standard. The packet flow process in software defined network is demonstrated in Figure 1.

Figure 1. Open flow packet flow chart Software defined network emulation with OpenFlow protocol (Tsehay Admassu Assegie)



ISSN: 2252-8814

RESEARCH METHOD To create a software defined networks, we have used mininet, emulation software defined for modeling and testing software defined network. 3.1. Creating SDN network To create model of a software defined network, we used a mininet, which is a very powerful LINUX based software defined network emulation tool used by many researchers. It is customizable tool and allows setting certain parameters like bandwidth, delay, packet loss and queue size to different links in OpenV Switches used in software defined network. To model a software defined network in an OpenFlow using mininet emulator we have followed the following steps: a.

Creating software defined networks using command: tt@ubuntu:~$ sudo mn –topo single,4 –controller remote

The command sudo mn –top=single, 1, 4 –controller=remote creates a software defined network with a controller an OpenFLow switch and four hosts. b.

Listing the nodes available in the software defined network using command nodes: mininet> nodes

The result of the command shows: available nodes are: c0 h1 h2 h3 h4 s1 The hosts h1, h2, h3 and h4 are created and a controller is added into the software defined network. In Figure 2, all the virtual hosts are connected to the open flow switches, and the open flow switches are connected to the POX controller. The POX controller is a platform implemented in python to emulate the control plane in software defined network. The OpenFlow switches are used to connect the hosts with the POX controller. In this topology the controller is configured with port 6633 and loopback address of 127.0.0.1.

Figure 2. A software defined network, hosts and OpenFlow switches connected in a linear topology 3.2. Managing OpenFlow switch We have started the POX controller with a Simple_Switch application. The switch application keeps tracks of where the host with MAC address is located and accordingly sends packets towards the destination and does not flood it out through all ports. The SDN switch (for instance, an OpenFlow switch), the SDN controller, and the interfaces present on the controller for communication with forwarding devices, Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 70 – 76

Int J Adv Appl Sci

ISSN: 2252-8814



generally southbound interface (OpenFlow) and network applications interface (northbound interface) are the fundamental building blocks of an SDN deployment. Switches in an SDN are often represented as basic forwarding hardware accessible via an open interface, as the control logic and algorithms are offloaded to a controller. OpenFlow switches come in two varieties: pure (OpenFlow-only) and hybrid (OpenFlow-enabled). OpenFlow switch is a basic forwarding element, which is accessible via OpenFlow protocol and interface. Although at first glance this setup would appear to simplify the switching hardware, flow-based SDN architectures such as OpenFlow may require additional forwarding table entries, buffer space, and statistical counters that are not very easy to implement in traditional switches with application specific ICs (ASICs). In an OpenFlow network, switches come in two flavors, hybrid (OpenFlow enabled) and pure (OpenFlow only). Hybrid switches support OpenFlow in addition to traditional operation and protocols (L2/L3 switching). An OpenFlow switch consists of a flow table, which performs packet lookup and forwarding. Each flow table in the switch holds a set of flow entries that consists of: Header fields or match fields, with information found in packet header, ingress port, and metadata, used to match incoming packets. Counters, used to collect statistics for the particular flow, such as number of received packets, number of bytes, and duration of the flow. A set of instructions or actions to be applied after a match that dictates how to handle matching packets. For instance, the action might be to forward a packet out to a specified port. OpenFlow Switch Packet processing logic: Create MAC table if (packet into switch) { parse packet to reveal src and dst MAC addr store in dictionary mapping between MAC and portl lookup dst MAC into port dictionary of switch to find next hop if(next hop is found) {crate flow mode send } else flood all ports=in_port Starting mininet with three hosts and one OpenFlow switch s1

Dump flow on switch 1

RESULT AND ANALYSIS In the experiment, we have used the topology of Mininet given in Figure 2. This topology includes two OpenFlow switch connected to four hosts a one OpenFlow reference controller. Upon execution of Mininet emulation environment with this topology, the OpenFlow controller and switch initiate Software defined network emulation with OpenFlow protocol (Tsehay Admassu Assegie)



ISSN: 2252-8814

the OpenFlow protocol, which can be captured and viewed in the Wireshark capturing window which is shown in Figure 3. On top of this emulated SDN platform, POX is used as the SDN controller. The following screenshot shows the captured traffic, which shows the Hello message, feature request/reply. This confirms that the OpenFlow switch in this setup is connected to the OpenFlow controller. Now we can check the connectivity of each host by a simple ping command: mininet> h1 ping –c 3 h2 and the result of the ping command request is demonstrated in Figure 4 and Figure 3 shows openflow captured packet.

Figure 3. OpenFlow traffic, captured in wireshark

Figure 4. The captured traffic after issuing an h1 ping –c 3 h2 command in mininet The ping sends three ping request packets as shown in Figure 6. A flow entry covering ICMP ping traffic was previously installed in the switch, so no control traffic was generated, and the packets immediately pass through the switch. An easier way to run this test is to use the Mininet CLI built-in pingall command, which does an all-pairs ping. Another useful test is a self-contained regression test. The following command created a minimal topology, started up the OpenFlow reference controller, runs an all-pairs-ping test, and tore down both the topology and the controller. The data rate of each emulated Ethernet link in Mininet is enforced by Linux Traffic Control (tc), which has a number of packet schedulers to shape traffic to a configured rate. Mininet allows you to set link parameters, and these can even be set automatically from the command line: sudo mn –link tc, bw=10, delay=10ms Check the traffic bandwidth using command and verify result:

This will set the bandwidth of the links to 10 Mbps and a delay of 10 ms. Given this delay value, the round trip time (RTT) should be about 40 ms, since the ICMP request traverses two links (one to the switch, one to the destination) and the ICMP reply traverses two links coming back. The round-trip time for packet is illustrated in Table 1. As shown in Table 1, the round trip time is higher in linear topology compared to the tee and hybrid topology. In Figure 5 the round trip time of each captured packet is shown and Figure 6 shows that, the tree topology in OpenFlow based software defined network performs better than the linear and hybrid topologies. The round-trip time for tree topology is much lower than all types of topologies used in OpenFlow based software defined networks. The hybrid approach performed moderately compared to the tree and linear topology whereas the linear topology is the lower in performance than the tree and hybrid topologies.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 70 – 76

Int J Adv Appl Sci



ISSN: 2252-8814

Figure 5. Round-trip delays between nodes for basic OpenFlow topologies

Figure 6. Round trip time between nodes, hybrid, linear and tree OpenFlow topologies Table 1. Round-trip delay between nodes for basic OpenFlow topologies # of packets 5 10 20 30 40 50

Hybrid topology (Round Trip Time in milliseconds) Maximum minimum 4.3 0.088 4.53 0.052 0.059 0.206 5.15 0.087 4.36 0.067 3.78 0.047

Tree topology (Round Trip Time in milliseconds) maximum minimum 13.3 0.09 0.375 0.079 0.317 0.063 0.305 0.037 0.188 0.056 2.95 0.079

Linear topology (Round Trip Time in milliseconds) maximum minimum 30.3 0.085 0.315 0.087 0.287 0.048 0.27 0.046 0.309 0.053 0.251 0.048

CONCLUSION Software defined network (SDN), which is often denoted as a revolutionary emergining idea in computer networking, promises to dramatically simplify network control, management, and enable innovation through network programmability. Mininet facilitates the creation and manipulation of software defined network components. The mininet is helpful to explore OpenFlow, which is an open interface for controlling the network elements through their forwarding tables. A network element can be converted into a switch or a router via low level primitives defined in the OpenFlow. In this study, we have emulated a software defined network using mininet and POX, a software platform developed by Python which we have used as a controller with OpenFlow switch and we have discussed the characteristics of software defined network. By issuing commands in the mininet environment we have showed that the network infrastructure can be virtualized for example Ethernet link bandwidth can be dynamically provisioned from a controller, like POX programmatically. Finally, the pefromance of three topologies namely, hybrid, tree and linear toplogy is evaluated in the emulated test bed environment and results shows that the hybrid toplogy is better in performance compared to the linear and tree topologies. REFERENCES [1] Faris Keti and Shavan Askar. “Emulation of Software Defined Networks Using Mininet in Different Simulation Environments,” 6th International Conference on Intelligent Systems, Modeling and Simulation, 2015. [2] Danda B. Rawat, “Software Defined Networking Architecture, Security and Energy Efficiency: A Survey,” IEEE Communication Surveys & Tutorials, Vol. 19, No. 1, 2017.

Software defined network emulation with OpenFlow protocol (Tsehay Admassu Assegie)



ISSN: 2252-8814

[3] Diego Kreutz, “Software-Defined Networking: A Comprehensive Survey,” IEEE, Vol. 103, No. 1, 2014. [4] Xuan-Nam Nguyen, Damien Saucez, Chadi Barakat, and Thierry Turletti, “Rules Placement Problem in OpenFlow Networks: A Survey,” IEEE Communication Survey and turotials, 2015. [5] Daniel King, Charalampos Rotsos, Alejandro Aguado, Nektarios Georgalas, and Victor Lopez, “The Software Defined Transport Network: Fundamentals, Findings and Futures,” IEEE, 2016. [6] Sakir Sezer, Sandra Scott-Hayward, and Pushpinder Chouhan, “Implementation Challenges for Software-Defined Networks,” IEEE Communications Magazine, July 2013. [7] Aashish Dugar and M Madiajagan, “Study on SDN using mininet,” Indian Journal of Computer Science and Engineering, 2016. [8] Rajiv Ranjan, Sumit Thakur, and Pankaj Rai, “Implementing the concept of software defined network in cloud computing,” International Journal of Computer Engineering and Applications, Vol. 11, No. 9, 2016. [9] Akhilesh Thyagaturu, Anu Mercian, Michael P. McGarry, Martin Reisslein, and Wolfgang Kellerer, “Software Defined Optical Networks (SDONs): A Comprehensive Survey,” IEEE, 2016. [10] K. Benzekki, A. El Fergougui, and A. Elbelrhiti Elalaoui, “Software-defined networking (SDN): A survey,” Secur. Commun. Netw., Vol. 9, No. 18, 2017. [11] Yogita Hande and M. Akkalakshmi, “A Study on Software Defined Networking,” International Journal of Innovative Research in Computer and Communication Engineering, Vol. 3, No. 11, 2015. [12] Wolfgang Braun and Michael Menth, “Software-Defined Networking Using OpenFlow: Protocols, Applications and Architectural Design Choices,” Future Internet, Vol. 6, No.2, 2014. [13] Zahra Al-Abri, Ahmed Al Maashri, Dawood Al-Abri, and Fahad Bait Shiginah, “Using SDN as a Technology Enabler for Distance Learning Applications,” Indonesian Journal of Electrical Engineering and Informatics (IJEEI), Vol. 6, No. 2, pp. 225-234, 2018. [14] Manar Jammala, Taranpreet Singha, Abdallah Shamia, RasoolAsalb, and Yiming Li, “Software-Defined Networking: State of the Art and Research Challenges,” Journal of Computer Networks, 2014. [15] R. Sahay, W. Meng, and Christian D. Jensen, “The application of Software Defined Networking on securing computer networks: A survey,” Journal of Network and Computer Applications, Vol. 131, pp. 89–108, 2019. [16] Sandhya, Yash Sinha, and K. Haribabu, “A survey: Hybrid SDN,” Journal of Network and Computer Applications, Vol. 100, pp. 35–55, 2017. [17] Murat Karakus and Arjan Durresi, “A survey: Control plane scalability issues and approaches in Software-Defined Networking (SDN),” Computer Networks, Vol. 112, pp. 279–293, 2017. [18] Tianzhu Zhang, Paolo Giaccone, Andrea Bianco, and Samuele De Domenico, “The role of the inter-controller consensus in the placement of distributed SDN controllers,” Computer Communications, Vol. 113, pp. 1–13, 2017. [19] Sibylle Schallera and Dave Hoodb, “Software defined networking architecture standardization,” Computer Standards & Interfaces, Vol. 54, pp. 197–202, 2017. [20] Franciscus Wibowo, Mark A. Gregory, Khandakar Ahmed, and Karina M. Gomez, “Multi-domain Software Defined Networking: Research status and challenges,” Computer Standards & Interfaces, Vol. 54, pp. 197–202, 2017. [21] D. Singh, Bryan Ng, Yuan-Cheng Lai, Ying-Dar Lin, and Winston Seah, “Modeling Software-Defined Networking: Software and hardware switches,” Journal of Network and Computer Applications, Vol. 122, pp. 24–36, 2018. [22] Celyn Birkinshaw, Elpida Rouka, and Vassilios G. Vassilakis, “Implementing intrusion detection and prevention system using software-defined networking: Defending against port-scanning and denial-of-service attacks,” Journal of Network and Computer Applications, March 2019. [23] Saleh Asadollahi, Bhargavi Goswami, and Atul M Gonsai, “Software Defined Network, Controller Comparison,” International Journal of Innovative Research in Computer and Communication Engineering, Vol. 5, Special Issue 2, April 2017. [24] Idris Zoher Bholebawa and Upena D. Dalal, “Design and Performance Analysis of OpenFlow-Enabled Network Topologies Using Mininet,” International Journal of Computer and Communication Engineering, 2016. [25] Tsehay Admassu Assegie and Pramod Sekharan Nair, “The performance of Gauss Markov’s mobility model in emulated software defined wireless mesh network,” Indonesian Journal of Electrical Engineering and Computer Science, Vol. 18, No. 1, April 2020.

Int J Adv Appl Sci, Vol. 9, No. 1, March 2020: 70 – 76

Institute of Advanced Engineering and Science Indonesia

: D2, Griya Ngoto Asri, Bangunharjo, Sewon, Yogyakarta 55187, Indonesia

Malaysia

: 51 Jalan TU 17, Taman Tasik Utama, 75450 Malacca, Malaysia

COPYRIGHT TRANSFER FORM (Please compile this form, sign and send by e-mail) Please complete and sign this form and send it back to us with the final version of your manuscript. It is required to obtain a written confirmation from authors in order to acquire copyrights for papers published in the International Journal of Advances in Applied

Sciences (IJAAS) Full Name and Title Organisation Address and postal code City Country Telephone/Fax E-mail Paper Title Authors

Copyright Transfer Statement The copyright to this article is transferred to Institute of Advanced Engineering and Science (IAES) if and when the article is accepted for publication. The undersigned hereby transfers any and all rights in and to the paper including without limitation all copyrights to IAES. The undersigned hereby represents and warrants that the paper is original and that he/she is the author of the paper, except for material that is clearly identified as to its original source, with permission notices from the copyright owners where required. The undersigned represents that he/she has the power and authority to make and execute this assignment. We declare that: 1. This paper has not been published in the same form elsewhere. 2. It will not be submitted anywhere else for publication prior to acceptance/rejection by this Journal. 3. A copyright permission is obtained for materials published elsewhere and which require this permission for reproduction. Furthermore, I/We hereby transfer the unlimited rights of publication of the above mentioned paper in whole to IAES. The copyright transfer covers the exclusive right to reproduce and distribute the article, including reprints, translations, photographic reproductions, microform, electronic form (offline, online) or any other reproductions of similar nature. The corresponding author signs for and accepts responsibility for releasing this material on behalf of any and all co-authors. This agreement is to be signed by at least one of the authors who have obtained the assent of the co-author(s) where applicable. After submission of this agreement signed by the corresponding author, changes of authorship or in the order of the authors listed will not be accepted.

Retained Rights/Terms and Conditions 1. 2.

Authors retain all proprietary rights in any process, procedure, or article of manufacture described in the Work. Authors may reproduce or authorize others to reproduce the Work or derivative works for the author’s personal use or for company use, provided that the source and the IAES copyright notice are indicated, the copies are not used in any way that implies IAES endorsement of a product or service of any employer, and the copies themselves are not offered for sale. Although authors are permitted to re-use all or portions of the Work in other works, this does not include granting third-party requests for reprinting, republishing, or other types of re-use.

Yours Sincerely,

Corresponding Author‘s Full Name and Signature Date: ……./……./…………

IAES Journals, http://iaescore.com/journals International Journal of Advances in Applied Sciences (IJAAS), email: iaes.editor@gmail.com, editor@iaesjournal.com

Guide for Authors International Journal of Advances in Applied Sciences (IJAAS) is a peer-reviewed and open access journal dedicated to publish significant research findings in the field of applied and theoretical sciences. The journal is designed to serve researchers, developers, professionals, graduate students and others interested in state-ofthe art research activities in applied science areas, which cover topics including: chemistry, physics, materials, nanoscience and nanotechnology, mathematics, statistics, geology and earth sciences. Papers are invited from anywhere in the world, and so authors are asked to ensure that sufficient context is provided for all readers to appreciate their contribution.

The Types of papers The Types of papers we publish. The types of papers that may be considered for inclusion are: 1) Original research; 2) Short communications and; 3) Review papers, which include meta-analysis and systematic review

How to submit your manuscript All manuscripts should be submitted online at http://iaescore.com/journals/index.php/IJAAS

General Guidelines 1) Use the IJAAS guide (http://iaescore.com/journals/ourfiles/IJAAS_guideforauthors_2019.docx) as template. 2) Ensure that each new paragraph is clearly indicated. Present tables and figure legends on separate pages at the end of the manuscript. 3) Number all pages consecutively. Manuscripts should also be spellchecked by the facility available in most good word-processing packages. 4) Extensive use of italics and emboldening within the text should be avoided. 5) Papers should be clear, precise and logical and should not normally exceed 3,000 words. 6) The Abstract should be informative and completely self-explanatory, provide a clear statement of the problem, the proposed approach or solution, and point out major findings and conclusions. The Abstract should be 150 to 250 words in length. The abstract should be written in the past tense. 7) The keyword list provides the opportunity to add keywords, used by the indexing and abstracting services, in addition to those already present in the title. Judicious use of keywords may increase the ease with which interested parties can locate our article. 8) The Introduction should provide a clear background, a clear statement of the problem, the relevant literature on the subject, the proposed approach or solution, and the new value of research which it is innovation. It should be understandable to colleagues from a broad range of scientific disciplines. 9) Explaining research chronological, including research design and research procedure. The description of the course of research should be supported references, so the explanation can be accepted scientifically. 10) Tables and Figures are presented center. 11) In the results and discussion section should be explained the results and at the same time is given the comprehensive discussion. 12) A good conclusion should provide a statement that what is expected, as stated in the "Introduction" section can ultimately result in "Results and Discussion" section, so there is compatibility. Moreover, it can also be added the prospect of the development of research results and application prospects of further studies into the next (based on the results and discussion). 13) References should be cited in text by numbering system (in Vancouver style), [1], [2] and so on. Only references cited in text should be listed at the end of the paper. One author should be designated as corresponding author and provide the following information: • E-mail address • Full postal address • Telephone and fax numbers Please note that any papers which fail to meet our requirements will be returned to the author for amendment. Only papers which are submitted in the correct style will be considered by the Editors.

International Journal of Advances in Applied Sciences (IJAAS) Institute of Advanced Engineering and Science (IAES) e-mail: ijaas@iaesjournal.com

IJAAS Journal Order Form Volume

Number

Amount

Price (Rp)

Price (USD)

Total

Name

Company

Address

City / State

Zip

Telephone/Fax

: ........................, ..........................

Signature: ..................................

Order form for subscription should be sent to the editorial office by fax or email

Payment by Bank Transfer Bank Account name (please be exact)/Beneficiary: LINA HANDAYANI Bank Name: CIMB NIAGA Bank Branch Office: Kusumanegara City: Yogyakarta Country :Indonesia Bank Account # : 5080104447117 Bank Swift : BNIAIDJAXXX >>> Please find the appropriate price in the price list on next page >>>

The price list for domestic and foreign subscribers

Volume

Number

Year

1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4 5 5 5 5 6 6 6 6 7 7 7 7 8 8 8 8 9 9 9 9

1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4

2012 2012 2012 2012 2013 2013 2013 2013 2014 2014 2014 2014 2015 2015 2015 2015 2016 2016 2016 2016 2017 2017 2017 2017 2018 2018 2018 2018 2019 2019 2019 2019 2020 2020 2020 2020

Price (IDR) for domestic subscribers 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00

Price (USD) for foreign subscribers 36 36 36 36 36 36 36 36 40 40 40 40 40 40 40 40 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44

The price included the printing, handling, packaging and postal delivery fees of the hardcopy to the address of the authors or subscribers (with Registered Mail). For foreign subscribers, an additional fee is charged if you would your order is mailed via Express Mail Service (EMS): - $25 for ASIA Continent - $35 for AUSTRALIA Continent - $35 for AFRICA Continent - $39 for AMERICA Continent - $39 for EUROPE Continent (No additional fee for delivering your order by Registered Mail).