		FeedBack \| Contact Us \| Links \| Site Map

Welcome to Journal of Theoretical and Applied Information Technology

Submit Paper / Call for Papers

Journal receives papers in continuous flow and we will consider articles from a wide range of Information Technology disciplines encompassing the most basic research to the most innovative technologies. Please submit your papers electronically to our submission system at http://jatit.org/submit_paper.php in an MSWord, Pdf or compatible format so that they may be evaluated for publication in the upcoming issue. This journal uses a blinded review process; please remember to include all your personal identifiable information in the manuscript before submitting it for review, we will edit the necessary information at our side. Submissions to JATIT should be full research / review papers (properly indicated below main title).

Journal of Theoretical and Applied Information Technology

October 2023 | Vol. 101 No.19

Title:	WILL TIKTOK USERS TAKE ADVANTAGE OF THE LIVE-SHOPPING FEATURE IN ONLINE SHOPPING?
Author:	MERLIN REINETA, VIANY UTAMI TJHIN
Abstract:	Live shopping is a breakthrough and has become a new trend in recent years. Many applications add live shopping features to be able to compete with their competitors and increase transactions made by users. TikTok is one of those apps with live streaming functionality. The researcher conducted this research to study the elements that influence the Buy Intention of consumers shopping directly on TikTok because this function is popularly discussed and used. This research was conducted using an online questionnaire containing 18 questions with a scale of 1-5. The results of the questionnaire obtained 110 valid data. After the data is collected, the research is continued using PLS-SEM based on the TAM model which has seven research hypotheses. The results of the analysis using the SmartPLS software show that Trust, information quality, convenience, usability, and Attitude have a positive effect on Purchase Intentions. After conducting the analysis, the researchers also offered suggestions and feedback to the TikTok business or application, especially for the live streaming function, to improve the quality of the feature according to the analysis findings. Some of the suggestions and feedback offered are: 1) Improving the security of personal data, transaction processing, and conformity of product information. 2) Improving the accuracy of information and conformity of information. 3) Improving convenience and usability by making the application interface more straightforward but with complete information. 4) Improving the quality of information by checking the quality of information.
Keywords:	Live Shopping, Purchase Intention, Consumer Attitudes, E-Commerce, Tiktok
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	EXPLORING STRATEGIES FOR OVERCOMING ISSUES OF USER INVOLVEMENT IN AGILE SOFTWARE DEVELOPMENT: A SYSTEMATIC LITERATURE REVIEW
Author:	QUDRATTULLAH OMERKHEL, OTHMAN MOHD YUSOP, SAIFUL ADLI ISMAIL, AZRI AZMI
Abstract:	The present systematic literature review (SLR) explores the challenges and strategies associated with managing users during requirement elicitation within agile software development. Drawing insights from an analysis of 24 relevant studies, this study comprehensively examines the issues that arise and the effective approaches to overcome them. The findings reveal five prominent challenges of user involvement during requirement elicitation. The most dominant issues identified are the lack of user involvement, insufficient user knowledge, and a deficit in the expertise of the Product Owner. These challenges can hinder the effective integration of user perspectives and needs into the development process. To address these challenges, the study identifies seven strategies that Product Owners can adopt to facilitate effective user involvement. These strategies include Mind Maps, User Interface Mockups, Workshops, Hybridism (combining agile and non-agile techniques), Face-to-Face Meetings, Continuous Delivery, and Training and Learning initiatives. The application of these strategies empowers Product Owners and software practitioners to enhance user involvement, improve communication, and streamline the requirement elicitation process in agile software development. The outcomes of this SLR provide valuable insights for both researchers and software practitioners, exploring the complex dynamics of user involvement in agile contexts. By recognizing these challenges and deploying effective strategies, software development teams can ensure more successful requirement elicitation processes, leading to the creation of software products that better align with user needs and expectations. This review contributes to a deeper understanding of user involvement challenges and offers actionable guidance for optimizing the requirement elicitation within agile software development paradigm.
Keywords:	User Involvement, Requirement Elicitation, Agile Software Development, Systematic Literature Review.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	REAL AND SIMULATED MASKED FACE RECOGNITION WITH A PRE-TRAINED MODEL
Author:	AUDREY ANAK ALBERT, SOO SEE CHAI, KOK LUONG GOH, KIM ON CHIN
Abstract:	Facial recognition has currently become indispensable owing to the efficacy of precise identification verification. Because of the distinctiveness of human biometrics, face recognition enables humans to communicate with technology while maintaining their privacy. Advancements in pre-trained models such as FaceNet have enabled improvement in identification accuracy in face recognition technology. Response to the Covid-19 pandemic has led to the replacement of conventional face recognition with masked face recognition. This change has encouraged the use of collaboration to resolve the related issues, which has resulted in the development of algorithms for face occlusion, collection of data on masked and unmasked faces and improvement of pre-trained models. Current research has utilised custom datasets or a specially produced dataset for masked face recognition. To increase the amount of data available for modelling, some studies have implemented mask simulation in facial photos. In this study, FaceNet is evaluated on two datasets: the real-masked face recognition dataset and the simulated masked face recognition dataset. Particularly, we highlight the performance of FaceNet on simulated masked faces. Using simulated masks achieved 67% accuracy, while the use of real masks achieved 84.3%. Results from the two datasets are compared with each other and with other studies using different pre-trained models with similar datasets. This study reveals that simulated masked faces perform less effectively than real masked faces, as corroborated by various other studies.
Keywords:	Masked face recognition, Face recognition, Pre-trained model, FaceNet
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	ANALYSIS FACTORS THAT AFFECTING CUSTOMER’S SATISFACTION IN PEER-TO-PEER LENDING APPLICATION USAGE
Author:	DANIEL KURNIAWAN, S.E., IR. TOGAR ALAM NAPITUPULU
Abstract:	The declining of personal income and the rising percentage of employee layoffs during pandemic caused people to scramble for any aid in their finances. This should trigger financial service enablers to change their conventional business patterns to a more sophisticated and more attainable version. One of the alterations that can be done is by utilizing available online features such as developing mobile applications as their business advantage. To successfully gain market exposure, Peer-to-Peer Lending applications must think of ways on how to have an edge compared to the rest. Therefore, they need to know what factors can increase their customer’s satisfaction. The purpose is to know if there are correlation between why people use certain applications and trust factors are related to why people keep coming back to the application. This research uses a combination of different variables that is obtained from UTAUT, UTAUT2, and ITM models. Research respondents consist of 131 people that filled online questionnaire which was distributed online. Data calculation is done through SmartPLS 4.0 application. The research studied Effort Expectancy, Social Influence, Facilitating Conditions, Hedonic Motivations, Price Value, Structural Assurance, Personal Propensity to Trust, Firm Reputation, and Application Type as its variables. Research’s findings are Social Influence, Hedonic Motivations, Structural Assurance, Personal Propensity to Trust, Firm Reputation, and Application Type variables have a positive effect on customer’s satisfaction in Peer-to-Peer Lending application usage, and Application Type moderation on Social Influence variable helps to also increase its satisfaction. To sum up, Social Influence, Hedonic Motivations, and Trust Variables are the variables Peer-to-Peer Lending Companies must look after for, since they are the most important and impactful variables that will steer the market towards them.
Keywords:	Peer-to-Peer Lending, UTAUT, UTAUT2, Customer Satisfaction, ITM
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	MODIFICATION OF QUANTUM ALGORITHMS FOR QUDITS WITH AN EVEN NUMBER OF STATES IN ORDER TO OPTIMIZE AND REPLACE OBSOLETE ALGORITHMS
Author:	LARISSA CHERCKESOVA, ELENA REVYAKINA, OLGA SAFARYAN, KIRILL LYASHENKO
Abstract:	In this article, a modification of Shor's algorithm for multiple 2n number systems of quantum computer was implemented. Work was carried out to study the number systems of quantum computers, the features of their use in quantum programming, as well as the simplification of computations by Shor's algorithm itself. The purpose of Shor's algorithm is to factorize any number in less time. The modification of Shor's algorithm developed in the course of the study makes it possible to simplify the calculations of the algorithm, to reduce the volume of circuits (schemes), at least to two digits of a number, which will allow getting rid of unnecessary calculations. This article is devoted to the operation of the algorithm in different number systems. Because working with real quantum computers is available to the narrow circle of researchers, the application of the modification in the emulator can affect the computational speed; in this case, there may be cases when the modification can work more accurately. The author's modification reduces the number of required qubits to 2, practically without reducing the performance of the algorithm and its execution time. Consequently, the cost of the quantum circuit itself also decreases several times (by 5–6 times, since not 12, but 2 qubits are required).
Keywords:	Digital Security, Information Security, Qubit, Factorization, Quantum Programming.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	SOPHISTICATED LION OPTIMIZATION BASED ROUTING PROTOCOL (SLORP) FOR CONGESTION AVOIDANCE IN MOBILE AD-HOC NETWORK
Author:	MRS.S.PREEMA, DR.M.THILAGU
Abstract:	The network that does not require infrastructure or central access is called Mobile Ad-hoc Networks (MANET). MANET has a wide range of applications because of its quick and flexible networking style. Routing protocols for MANETs are often developed on the premise that all participating nodes are completely cooperative. Developing effective routing protocols for MANETs is a difficult endeavor. There has been great interest in algorithms inspired by swarm intelligence because they can provide optimal solutions with cheap cost, flexibility and high resilience. Furthermore, MANETs can deal with complex, large-scale issues without the aid of a centralized authority. A bad route in MANET can fail at any time, causing retransmissions and congestion on the network, significantly impacting performance. Sophisticated Lion Optimization-based Routing Protocol (SLORP) is a new bio-inspired routing protocol proposed in this paper for avoiding congestion in MANET by finding a stable route. The natural characteristics of the lion inspire SLORP. The five different stages of SLORP are (i) Pride Generation, (ii) Reproduction, (iii) Bonding, (iv) Handling, and (v) Termination. The delay and energy consumption are minimized in SLORP via improved routing synchronization with the network’s nodes. An evaluation of SLORP is performed using common metrics in NS3. SLORP outperforms existing routing protocols in terms of performance.
Keywords:	MANET, Routing, PSO, Optimization, Energy, Swarming.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	A HYBRID APPROACH FOR LANDMARK DETECTION OF 3D FACES FOR FORENSIC INVESTIGATION
Author:	SINCY JOHN, AJIT DANTI
Abstract:	Facial landmark detection is a key technology in many forensic applications, such as facial identification and facial reconstruction. However, the accuracy of facial landmark detection is often limited in 3D face images due to the challenges of occlusion, illumination, and pose variations. This paper proposes a hybrid approach for landmark detection of 3D faces for forensic investigation. A hybrid method of edge contour detection and Harris corner detection is proposed for feature extraction in face images for forensic investigation. Edge contour detection is used to detect the boundaries of the face, while Harris corner detection is used to detect the corners. The advantage of using a hybrid method of edge contour detection and Harris corner detection for feature extraction in face images is that it can capture both global and local features of the face. Edge contour detection can capture global features, such as the overall shape and outline of the face, while Harris corner detection can capture local features, such as the corners of the mouth, nose and eyes which are vital for facial reconstruction. Experimental results show that the proposed method outperforms existing landmark detection algorithms in terms of time complexity and minimum loss.
Keywords:	Facial Reconstruction, Edge Contouring Detection, Harris Corner Detection, Hybrid Method, Landmark detection.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	DEEP LEARNED KERNEL SPECTRAL CLUSTERING AND PEARSON RANK SWAPPING ANONYMIZATION FOR PRIVACY PRESERVED DATA PUBLISHING
Author:	A. MALAISAMY, DR. G. M. KADHAR NAWAZ
Abstract:	Privacy-preservation is a challenging issue with the increasing volumes of published data. To overcome such issues in data publishing, different methods of reduced risks related to published data have been developed. However, it has a vital problem to protect the users’ sensitive data in a publication. Since the attacker may hack the user data. Therefore, the Deep Learned Kernel Spectral Clustering-based Pearson Rank Proximity Swapping Anonymization (DLKSC-PRPSA) technique is developed for improving the data privacy preservation rate with lesser information loss. The proposed DLKSC-PRPSA technique collects the number of records from the dataset. Then the DLKSC-PRPSA technique trained the input records with several layers namely the input layer, two hidden layers, and the output layer. The numbers of records are given to the input layer of deep neural learning. Then the input is fed into the first hidden layer to group the records into different clusters using a radial basis kernelized spectral clustering technique. The clustered results are transformed into the next hidden layer. In the second hidden layer, the Pearson rank proximity swapping anonymization method is applied for interchanging the value of sensitive attributes to protect the original information. Finally, the anonymized results are obtained at the output layer for further processing. Experimental evaluation is carried out with adult datasets using different metrics such as privacy preservation rate, anonymity level, information loss, and time complexity. The experimental result confirms that the DLKSC-PRPSA technique efficiently increases the privacy preservation rate, anonymity level and minimizes the time complexity as well as information loss of data anonymization than the state-of-the-art methods.
Keywords:	Privacy Preservation, Data Publishing, Deep Learning, Radial Basis Kernelized Spectral Clustering, Pearson Rank Proximity Swapping Anonymization Method
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	PERFORMANCE ATTRIBUTES ANALYSIS OF NHPP-BASED SOFTWARE DEVELOPMENT COST MODEL WITH INVERSE-TYPE DISTRIBUTION PROPERTIES
Author:	SEUNG KYU PARK
Abstract:	In this study, after applying the Inverse-type distribution (Inverse-Exponential, Inverse-Rayleigh), which is known to be suitable for reliability research because it can explain various types of life distribution, to the NHPP-based software development cost model, and the attributes that determine the performance of the model were analyzed. Also, to evaluate the efficiency of the proposed model, the optimal model compared with the Goel-Okumoto basic model was also presented. Using the randomly collected failure time data, software failure phenomena were identified and applied to attribute analysis, and maximum likelihood estimation (MLE) was used for the solution of parameters. In conclusion, first, as a result of analyzing the properties of m(t) that affect development cost, the Inverse-exponential model and the Goel-Okumoto basic model were efficient with small prediction errors for the true value. Second, as a result of analyzing the properties of release time along with development cost, the performance of the Inverse-Rayleigh model was the best. Third, as a result of comprehensively evaluating the performance attributes (m(t), cost, release time) of the cost model presented in this work, it was confirmed that the Inverse-Rayleigh model was the best. Therefore, if software developers can efficiently utilize this data in the early process, it is expected that they will be able to efficiently explore and analyze the attributes that affect development cost performance.
Keywords:	Goel-Okumoto, Inverse-Exponential, Inverse-Rayleigh, NHPP, Performance Attributes, Software Development Cost Model
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	IDENTIFICATION OF CELL MEMBRANES IN 2D IMAGES USING COMPUTER VISION
Author:	PAUL D. SARAVIA-VELASQUEZ, ROXANA FLORES-QUISPE, YUBER VELAZCO-PAREDES
Abstract:	Today in many types of research the cellular structure has been studied to identify diseases or to do clinical diagnostics; of fungi and epithelial cells, because the increase in invasive fungal infections in recent years, especially in immunocompromised patients, has prompted the search for new antifungal agents with greater efficacy. Where the fungal cell membrane is enriched with various lipids belonging to the class of glycerophospholipids, sphingolipids, and sterols. In addition, the precise roles of membrane lipids in the organization of these membrane domains in epithelial cells, where they are polarized and maintain apical and basolateral membranes, are largely unknown. These epithelial cells have morphologically distinct membrane structures with specific functions. This research presents a significant contribution to achieving the precise identification of cell membranes in 2D images based on Inception-type CNN architecture specifically designed for this research context, allowing effective detection of cell membranes. In addition, Canny, Thresholding, and Gaussian noise filters were implemented to improve the edge detection extracting relevant information from the object of interest and reducing unnecessary elements in the image. In addition, other segmentation techniques and size adjustments were used to improve the variability present in the images. Finally, the experimental results show the best performance with an accuracy rate of 86.6% using the Adagrad optimizer, which proved its efficiency in the search for membranes in 2D images. Our proposal, based on CNN, image processing, and adjustment techniques, offers a robust approach to this task, strengthening research in this field and opening opportunities for future improvements.
Keywords:	Computer Vision, CNN, Cell Membranes
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	THE EFFECT OF GAMIFICATION ON CUSTOMER ENGAGEMENT IN E-COMMERCE
Author:	JEREMY HERMON HERMAWAN, VIANY UTAMI TJHIN
Abstract:	The purpose of this study is to determine the effect of using gamification features in e-commerce towards customer engagement. With the rapid growth of e-commerce in Indonesia, many e-commerce has implemented gamification in hopes of staying ahead of its competitor. Gamification is seen to be an alternative to attract customer and increase customer engagement. This research aims to learn the effect of gamification on customer engagement towards users that have used the gamification feature on e-commerce. This study used a quantitative method, and data had been obtained through questionnaires. In this research customer engagement were constructed by four variables, customer lifetime value, customer influence value, customer referral value, and customer feedback value. There were 114 valid data respondents, and were analysed using Smart PLS. The results stated that gamification had a positive effect towards four variables of customer engagement. Furthermore, implication had been provided and suggested companies, if to build a gamification feature, to focus on customer influence value and customer knowledge value.
Keywords:	E-Commerce, Gamification, Customer Engagement, Customer Influence, Customer Knowledge
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	FOUNDATION OF CHAOTIC MAPS BASED ON DIMENSIONS WITH RELATION TO THE PROPERTY OF CRYPTOGRAPHY AND THE MATHEMATICAL EXPRESSIONS: A SYSTEMATIC REVIEW
Author:	YAHAYA GARBA SHAWAI, MOHAMAD AFENDEE MOHAMED, USMAN HARUNA, MOHAMMED AMIN ALMAIAH,ABDALWALI LUFTI, SULAIMAN IBRAHIM MUHAMMAD
Abstract:	The aim of the review was to carry out a survey on chaotic system in relation to cryptography based on both confusion and diffusion properties, with regards to the mathematical expressions. The mathematical expressions were discussed with the respect to the dimension of the chaotic system. The review enable the author to investigate some existing survey’s within the relevant field, with aid of a new proposed systematic review framework known on as YAFSU. The framework considers the search strategy (search and study selection) with regards to the extracted data and synthesis implementation. It is assumed that, the review may be able to assist researcher’s with research interest on image encryption based on chaotic system, to discover the chaotic maps that may be applicable for the image encryption schemes formulations. The present survey limits it review on the basic relationship of chaotic system and cryptography, foundation of chaotic map based on the dimension with regards to the mathematical expression. It was recommended that empirical review should be undergone to take a comparative study for encryption algorithms; security and performance analysis; the elements of the cryptosystem. Moreover, discussions on chaotic maps should go beyond five dimensional systems, as higher dimensional chaotic systems were considered to be in to existence.
Keywords:	Information, Chaotic System, Cryptography, Encryption, Decryption
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	CUSTOMER RELATIONSHIP MANAGEMENT: TWO DATASET COMPARISON IN PERSPECTIVE OF BANK LOAN APPROVAL USING MACHINE LEARNING TECHNIQUES
Author:	DURGESH KUMAR SINGH, NOOPUR GOEL
Abstract:	One of the main goals of any banking industry that wants to last for a long time is to become profitable. Understanding the customer is necessary for providing services and products to the customer in accordance with his preferences and requirements. Client division and profiling are essential in accomplishing two principal goals of CRM (Customer Relationship Management) i.e.; client maintenance and client advancement. Under CRM, loans are the most important product that banks and other financial institutions offer to meet customers' needs, but determining which customers are eligible for loans is a significant challenge. However, loans come with a risk of default, or the possibility that some borrowers will not be able to repay the loans they have been given. As a result, banks that have a lot of non-performing loans may go bankrupt or become unstable as a result. The progression of innovation like artificial intelligence (AI), getting helpful data from client information is of central significance in nowadays. By developing, comparing, and testing the accuracy of various models using datasets from two banks, we contribute to commercial banks' efforts to predict borrowers' behaviors. Throughout the manuscript, the first dataset is referred to as Dataset-1, and the second dataset is referred to as Dataset-2. In order to determine which machine learning method is most effective for predicting bank loan default, base learners, ensemble, and voting are used. The results demonstrate that Random forest (RF) outperformed all other classifiers. Precision, recall, the f1-score, and the ROC (AUC) curve all supported the classifiers' findings. Because it saves both time and money, I would suggest that financial institutions employ machine learning methods.
Keywords:	CRM, Bank loan, Artificial intelligence, Ensemble, Voting, Recall, ROC (AUC).
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	EMBEDDED FPGA HARDWARE IMPLEMENTATION OF A PREPROCESSING ALGORITHM FOR SURVEILLANCE IMAGES
Author:	ISSAM BOUGANSSA, HICHAM BENRADI, ADIL SALBI, MOHAMED SBIHI, ABDELALI LASFAR
Abstract:	Surveillance images have become essential for deciphering many accidents and offenses, however, the major obstacle in this area is their quality, which is often obtained via low-resolution surveillance cameras, or in difficult light conditions. This work aims to implement in real-time on an FPGA-type embedded system, pre-processing algorithms to improve the quality of surveillance images by normalizing and enriching histograms, followed by edge detection algorithms, using software and hardware tools. The objective of this improvement is to allow professionals to dissect the necessary information more easily. To do this, it was decided to program the algorithms with a VHDL-type hardware description language and test them on more efficient and faster tools called Xilinx System Generator (XSG) which allows several algorithms to be tested on software, before implementation. This implementation is done on a XILINX SPARTAN-6 FPGA board using the ISE Design Suite tool from Xilinx.
Keywords:	Pre-Processing, Normalization, FPGA, Edge-Detection, VHDL.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	THE COUNTERMEASURES OF WIRELESS SENSOR NETWORK THREATS IN IOT SYSTEM
Author:	JAWAHER ALSHEHRI, ALMAHA ALHAMED, MOUNIR FRIKHA
Abstract:	Wireless sensor networks (WSNs) are a critical component in Internet of Things (IoT) systems, playing a crucial role in collecting and transmitting data. Wireless sensor networks (WSNs) often transmit sensitive data, such as personal information or industrial data. Without any proper protection, this transmitted data can be intercepted and exploited by malicious actors that use threats and attacks to achieve their goals. This paper explores the different types of threats faced by WSNs and proposes countermeasures to mitigate these threats. Through the implementation of these countermeasures, the security of WSNs in IoT systems can be strengthened, ensuring the safe and reliable transmission of data.
Keywords:	Wireless Sensor Networks, Threats, Attacks , Countermeasures, WSN Security, Iot Security
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	A HYBRID GENETIC BASED GREY WOLF OPTIMIZED SOPHISTICATED SUPPORT VECTOR MACHINE (SSVM) MODEL FOR SOFTWARE DEFECT PREDICTION
Author:	DR MEDHUNHASHINI, Dr. KS JEEN MARSELINE
Abstract:	Software Defect Prediction is one of the promising fields in software engineering, focusing on identifying and predicting the defective module in software before the testing phase begins. It helps to allocate resources in the testing phase cost-effectively. Developing a machine learning model that classifies the faulty module from non-faulty seems challenging. This paper focuses on developing an ensemble machine learning model, a Sophisticated Support Vector Machine (SSVM), for effective defect prediction. SSVM is built with the hybrid power of GA and GWO over the SVM. An enhanced Genetic Algorithm (GA) is used to select appropriate features from the defect dataset by Crossover of selected features. Grey Wolf Optimization (GWO) has been adopted to tune the hyperparameter of SVM's Radial Basis Function Kernel. The defect dataset JDT and MyLyn from the AEEEM repository is taken for experimentation. The model is investigated with 10-fold cross-validation, and performance is evaluated with a confusion matrix and F1 score. The results show the SSVM model classifies the defective from the non-defective module with an accuracy of 75.30 % and 77.30 %.
Keywords:	Quality, Software Defect, Genetic Algorithm, Grey Wolf Optimization, Support Vector Machine
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	A LEARNING BASED OPTIMIZED HYBRID MODEL FOR EFFICIENT AND SCALABLE LONG DOCUMENT CLASSIFICATION
Author:	AYESHA MARIYAM, SK. ALTAF HUSSAIN BASHA, S. VISWANADHA RAJU
Abstract:	There is exponential growth of text documents over cloud and Internet based storage infrastructures. There are many applications in the real world that exploit such documents. One such application is document classification which attracts academia and researchers in industry. However, classification of long length documents is found to be not trivial and complex phenomenon. Traditional approaches in deep learning could work on them with full information availability for feature representation leading to deteriorated performance. There were efforts by researchers combining multiple deep learning approaches to realize simplified feature representation of long documents. However, they suffer from issues with hyper parameter optimization and consumption of more resources. To address these problems, in this paper, we proposed a framework known as Long Document Classification Framework (LDCF) which exploits multiple deep learning models appropriately besides enhancing them with hyper-parameter optimization. It is an Artificial Intelligence (AI) enabled approach for scalable solution. Our empirical study has revealed that random search based approach for hyper-parameter optimization could improve the performance of the framework. We proposed an algorithm known as Learning based Optimized Hybrid Model for Long Document Classification (LOHM-LDC) to realize LDCF. It exploits Deep Reinforcement Learning (DRL) to pick text blocks intelligently. Our empirical study with ArXiv dataset, consisting of 1.7 million long documents, has revealed that our hybrid model has potential to outperform existing models with 96.67% accuracy.
Keywords:	Deep Learning, Artificial Intelligence, Hyper-Parameter Optimization, Long Document Classification, Hybrid Deep Learning Model
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	MULTI LINEAR REGRESSION-BASED IOT AND FOG COMPUTING ON MAINTENANCE PREDICTIONS APPROACH FOR EFFICIENT ASSET MANAGEMENT IN INDUSTRY REVOLUTION 4.0
Author:	RAHUL KUMAR SINGH, SHOKHJAKHON ABDUFATTOKHOV, DR SANJIV RAO GODLA, DR. VUDA SREENIVASA RAO, PROF. TS. DR. YOUSEF A. BAKER EL-EBIARY, RICARDO FERNANDO COSIO BORDA
Abstract:	Industry 4.0 makes it possible for new developments in technology like Big Data Analytics and Machine Learning to be successfully incorporated into and combined with current production processes, allowing smart manufacturing. With the use of predictive maintenance, a company owner can make decisions like replacing or fixing a component before it fails and affects the entire production line. To optimize work distributions and maintenance prediction models, Industry 4.0 (I4.0) necessitates good asset management. The Multi Linear Regression (MLR) based predictive maintenance in IoT and fog computing is presented in this study. The data produced by I4.0's Industrial Internet of Things (IIoT) enables information transparency and process management. Regular maintenance enables the business manager to make choices like when to fix or substitute a part before it malfunctions and disrupts the whole manufacturing process. Thus the study illustrates a forecasting model for predicting rapid failure in industrial machinery and to enable an effective production and servicing process. The outcome demonstrates that the suggested solution performs better than current methods in terms of computational cost execution time, and energy usage. In comparison to the second-best outcomes, the execution period is quicker, the cost is less, and the amount of energy consumption is lower.
Keywords:	Internet of Things (IoT); Fog Computing; Industry 4.0; Asset Management; Multi Linear Regression (MLR)
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	BLOCKCHAIN DATA PRIVACY SOLUTION BASED ON IPFS CRYPTOGRAPHIC PROTOCOLS
Author:	ABDELAZIZ ELBAGHDADI, ASMAA HILMI, SOUFIANE MEZROUI, AHMED EL
Abstract:	To store the data in a decentralized and distributed way without third entity the block chain technology is used. The data in this technology is immutable and verifiable and its works without a third party. But blockchains faces of problems of scalability, security, and potential privacy, such as the ability to link transactions, the privacy of data on the blockchain or the conformity with privacy regulations. In this work, we propose an approach to share images with the help of blockchain, Interplanetary File System (IPFS) Protocol and visual cryptography. Furthermore, the proposed architecture aimed to solve the problem of image data privacy in distributed system. The proposed approach proceeds to encrypt and decrypt images using the RGB color decomposition principle (Red, Green, and Blue). Then the blockchain and IPFS protocol are used to ensure the data privacy in the distributed system. The simulation results of the proposed solution provide high security, and the security analysis including Peak Signal to Noise Ratio (PSNR), histogram analysis and correlation coefficient maintains a lossless encryption and reveals a high performance.
Keywords:	Blockchain, Cryptography, Security, Data Privacy, Confidentiality
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	FOREST FIRE DETECTION AND MONITORING THROUGH ENVIRONMENT SOUND SPECTRUM USING DEEP LEARNING
Author:	MOUNIR GRARI, MOHAMMED BOUKABOUS, MIMOUN YANDOUZI, MOHAMMED BERRAHAL, IDRISS IDRISSI, OMAR MOUSSAOUI, MOSTAFA AZIZI, MIMOUN MOUSSAOUI
Abstract:	Forests are one of the most important ecosystems on Earth. They play a vital role in regulating the climate and act as a renewable source of air for human beings. However, forests are really threatened by fires. When wildfires occur outside of their natural range or size, they can become a real danger to life and property. In this paper, we propose an original novel approach for detecting forest wildfires, based on collected data of fire sounds. This method employs a deep learning (DL) model to analyze and classify environment sounds into two classes: “Fire” or “No fire” (usual forest sounds). The model must first be trained on a set of environmental sounds in order to learn and identify fire sound patterns from other sounds. With this model, we achieve an impressive accuracy of 94.24% on the testing sub-dataset. Notably, the model consists of only 1789 parameters, rendering it exceptionally lightweight. This quality makes it highly conducive for deployment across various platforms such as IoT devices, embedded systems, or mobile devices. Integrating this model into forest environments and fortifying it with complementary tools for comprehensive validation could enable us to promptly notify decision-makers or relevant authorities, facilitating timely and decisive actions.
Keywords:	Forest Fires, Wildfires, Deep Learning (DL), Sound spectrum, Audio environment
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	ANALYSIS OF MULTICLASS IMBALANCE HANDLING IN RED WINE QUALITY DATASET USING OVERSAMPLING AND MACHINE LEARNING TECHNIQUES
Author:	UMA RANI V, KALADEVI R, JEBAMALAR TAMILSELVI J, SARASU P, CHARLES PRABU V
Abstract:	Wine quality is very important in the wine industry and is determined by its features and flavors. The primary goal of this study is to balance the wine quality data by generating synthetic data and using a machine learning model to predict. For the current research, a multiclass unbalanced red wine quality data set is obtained from UCI resources. To balance the multiclass imbalance red wine quality data set, SMOTE and its six derivatives, including SMOTENC, SMOTE EN, SMOTE Tomek, SVM-SMOTE, Borderline SMOTE, and SMOTE ENN, are used. To predict red wine quality, seven machine-learning approaches, including Logistic Regression, Decision Tree, Random Forest, Extra Tree, XG-Boost, AdaBoost, and Bagging classifier, were trained and assessed. According to the results of this experiment, a combination of SMOTE ENN+ ETC has a higher precision of 0.96, recall of 0.96, specificity of 0.99, f1 score of 0.96, geometric mean of 0.97, and indexed balance score of 0.95 than all other SMOTE variations. SMOTE ENN + ETC has a higher accuracy of 95.8% when compared to other models. As a result, this combination is utilized to forecast red wine quality.
Keywords:	Wine Quality, Multiclass imbalance, SMOTE and its Variants, Oversampling Machine Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	HIGH GAIN NARROW BAND SMALL SIGNAL HYBRID UNIT DARLINGTON PAIR AMPLIFIER
Author:	PRATIMA SONI, GEETIKA SRIVASTAVA, SYED SHAMROZ ARSHAD, ADITYA, SACHCHIDANAND SHUKLA
Abstract:	Popularity of smart home gadgets gives rise to the demand for enhanced (wireless) RF power amplifiers. The global market of RF power amplifiers registered a CAGR of 3.2% between 2017 and 2021 and is expected to exhibit a growth of 15% from 2021 to 2032. The rising need of such RF amplifiers opens the path to a wide research in this field to fulfill the requirements. The Darlington pair amplifiers having high current gain still have the issue of poor response at higher frequencies, which can be removed by insertion of load at collector in the reference amplifier circuit to achieve higher voltage gain, power gain, bandwidth, slew rate and lower values of noise and harmonic distortion .A unique hybrid unit Darlington pair amplifier circuit with BJT-MOS is proposed in the present paper whose maximum voltage gain of 281.331, current gain of 33.698, bandwidth of 22.562 MHz, total harmonic distortion of 5.2*10-06 and slew rate of 106.389 V/m Sec and can operate on input signal in the range of 1mV – 10mV.The result shows that proposed amplifier has variety of applications may be used as pre amplifiers in VLF receivers, ultrasonic transmitters and, wireless communication, radio receivers, defense avionics, etc.
Keywords:	Darlington Amplifiers, MOSFET Amplifiers, Small–Signal Amplifiers, High Gain Amplifiers, High Frequency Amplifiers, Hybrid Configuration
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	AUGMENTED DOLPHIN SWARM OPTIMIZATION-BASED SECURED GAUSSIAN AD-HOC ON-DEMAND DISTANCE VECTOR (ADSO-SGAODV) ROUTING FOR QUALITY OF SERVICE ENHANCEMENT IN MOBILITY-ENABLED WIRELESS SENSOR NETWORK
Author:	V.VEERAKUMARAN, Dr. ARUCHAMY RAJINI
Abstract:	Mobility-Enabled Wireless Sensor Networks (ME-WSNs) are a specialized class of wireless sensor networks (WSNs) that incorporate the capability of node mobility. Node mobility in ME-WSNs introduces several advantages but brings new challenges in efficiently routing data packets and allowing intruders to join the network. To address the routing challenges in ME-WSNs and enhance Quality of Service (QoS), an “Augmented Dolphin Swarm Optimization-Based Secured Gaussian Ad-Hoc On-Demand Distance Vector (ADSO-SGAODV)” routing protocol is proposed. ADSO-SGAODV works by efficiently discovering and maintaining routes for data transmission while conserving energy and ensuring secure communication. It employs a hybrid optimization-based approach, combining Gaussian AODV and Dolphin Swarm Optimization (DSO). ADSO-SGAODV utilizes a Support Vector Machine (SVM) to ensure intelligent decision-making in route selection. ADSO-SGAODV selects cluster heads based on fuzzy logic and specific criteria such as energy level and distance to the base station. This feature enhances network scalability and load balancing, ensuring efficient utilization of resources in a dynamic ME-WSN environment. ADSO-SGAODV focuses on providing robust security measures to safeguard data during transmission. Secure Communication protocols are implemented to encrypt data, preventing unauthorized access and maintaining data confidentiality. Trust-Based Access Control with Encryption is employed in ADSO-SGAODV to establish trust among nodes and ensure data communication integrity within the network. Through extensive simulations in various scenarios, ADSO-SGAODV has demonstrated a superior packet delivery ratio, throughput, energy consumption, and adaptability to node mobility. The protocol’s intelligent and energy-efficient working mechanism makes it a promising solution for enhancing QoS in ME-WSNs, addressing the unique challenges posed by mobility, and ensuring reliable and secure data transmission in dynamic environments.
Keywords:	QoS, AODV, Gaussian, Routing, Security, WSN
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	A MACHINE LEARNING-BASED CLASSIFIER FOR ANTICIPATING RISK FACTORS ASSOCIATED WITH CERVICAL CANCER
Author:	ANUSHA R, DR SRINIVAS PRASAD
Abstract:	Cervical cancer ranks as the second most prevalent form of cancer among women, but it is also highly preventable. Numerous studies have underscored the widespread lack of knowledge surrounding cervical cancer and its prevention. In developing countries, medical students represent the future healthcare professionals who can play a pivotal role in increasing public awareness and assessing knowledge about symptoms and risk factors. The present research focused on investigating the causes of cervical cancer and preventive methods for women who have already been diagnosed with this life-threatening disease. Each year, a significant number of women receive a diagnosis of cervical cancer, resulting in a substantial loss of lives worldwide. The Human papillomavirus (HPV) has been recognized as a significant risk factor for cervical cancer, and fortunately, effective prevention measures are available to a large extent. However, most cervical cancer cases are detected in economically disadvantaged countries where organized HPV screening or vaccination programs are lacking. High-income countries that have implemented comprehensive screening programs have successfully reduced the prevalence and mortality rates of cervical cancer by 50% over the past three decades. Fertility-preserving surgical approaches are now considered the standard treatment for women diagnosed with initial-stage, mild-risk cervical cancer. As cervical cancer remains a significant health concern for women, implementing a comprehensive strategy for prevention and control is paramount in our efforts to eliminate this disease. This research aims to introduce an ensemble-based classifier that enhances the early detection and prediction of cervical cancer. Through comparative analysis with other base classifiers, our proposed classifier demonstrates superior accuracy and performance.
Keywords:	Cervical Cancer, Screening, Risk Factors, HPV, Ensembled Enabled Classifier
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	E-PORTFOLIO FOR DIGITAL UNIVERSITIES USING SMART CONTRACTS ON INTELLIGENCE BLOCKCHAIN TECHNOLOGY
Author:	PINYAPHAT TASATANATTAKOOL, KATEKEAW PRADIT, PALLOP PIRIYASURAWONG
Abstract:	This research aims to develop and evaluate an e-portfolio for digital universities using smart contracts on intelligence blockchain technology. The analysis was carried out using synthesis, design, and evaluation in accordance with a conceptual framework and process consisting of four parts: Part 1: Input factors, which are objectives, users, equipment, and content. Part 2: The workflow, including interoperable technologies such as smart contracts, blockchain, artificial intelligence, and electronic portfolio formats. Part 3: Expert evaluation of the portfolio based on the research objectives. Part 4: The response to feedback from an expert suitability assessment. The suitability for digital universities of electronic portfolios using smart contracts on intelligence blockchain technology was found to be very good. Consequently, the model that has been developed may be an efficient platform to lead to a digital university according to the mission set by the university. This model serves as a framework for the development of various electronic performance evaluation methods, including but not limited to assessing employee performance and determining wage increments. These methods utilize data obtained from stored electronic records.
Keywords:	E-Portfolio, Smart Contracts, Blockchain Technology, Artificial Intelligence, Digital University
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	QUANTIFYING THE IMPACT OF WEARABLE HEALTH MONITORING AND MACHINE LEARNING RESEARCH: A BIBLIOMETRIC ANALYSIS
Author:	JEENA JOSEPH, JOBIN JOSE, DEUMY JOHN, SREENA V NAIR
Abstract:	Wearable devices that monitor various aspects of an individual's health in real-time are becoming more prevalent. Machine learning algorithms can analyze this medical information resulting in early disease prediction, customized treatment strategies, and enhanced healthcare delivery. Advancements in electronics and machine learning techniques are continuously expanding the potential for developing wearable, miniaturized, and, more particularly, biomedical sensor devices that can integrate sufficient cognitive capacity to analyze captured signals and respond to them. The study focuses on 582 articles published in the Scopus database between 1999 and 2023, highlighting the significant features, including co-authorship, publication patterns, word frequency, co-citation analysis, bibliographic relationships, and much more using VOSviewer and Biblioshiny. Overall, the study's findings provide insights into the current state of research in wearable health monitoring and machine learning and potential recommendations for future research in this rapidly emerging field.
Keywords:	Bibliometric Analysis, Wearable Health Monitoring, Machine Learning, VOSviewer, Biblioshiny.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	A HIGHLY EFFICIENT SYSTEM BASED ON DCT-TPLBP AND DCT-FPLBP DESCRIPTORS FOR FACE RECOGNITION
Author:	ELMAHDI BARRAH
Abstract:	In this paper, we describe an application for face recognition that combines local and global descriptors to improve performance. Here's a summary of the approach described in the paper: The performance of these local descriptors is comparatively better than global descriptors that operate on the entire image. To address this, the proposed approach applies local descriptors by dividing the image into blocks. By doing so, they aim to capture both the advantages of global and local methods. The local descriptors are applied in the DCT domain. The DCT is a commonly used transform technique that represents the image in terms of frequency components. By using the DCT domain, the aim is to exploit the frequency characteristics of facial features and capture relevant information for recognition. The proposed approach claims to provide a good compromise between global and local methods in terms of simplifying calculations while maintaining classification performance. This implies that the approach aims to strike a balance between accuracy and computational efficiency. Finally, we compare the results obtained from our approach with other local and global conventional approaches. The specific methods compared and the performance metrics evaluated should be detailed in the paper.
Keywords:	Face Detection, Face Recognition, Discrete Cosine Transforms (DCT), FPLBP, TPLBP.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	SENTIMENT ANALYSIS BASED ON 7P MARKETING MIX ASPECTS OF THE INDRIVER APPLICATION SERVICE USING THE BERT ALGORITHM, BASED ON USER REVIEWS ON THE GOOGLE PLAY STORE
Author:	HANDRIZAL, T. HENNY FEBRIANA HARUMY, HERRIYANCE, M. AMIRUL ILMI
Abstract:	This research aims to perform sentiment analysis on user reviews of the InDriver application service on the Google Play Store, focusing on the 7P marketing mix aspects. The analysis utilizes the BERT (Bidirectional Encoder Representations from Transformers) algorithm, known for its contextual text understanding and rich representation generation capabilities. The sentiment classification includes positive, negative, and neutral sentiments. The study collects data through scraping with the Google Play Scraper, followed by preprocessing steps such as tokenization and normalization. The collected dataset consists of 3028 user reviews. Three experiments are conducted, varying hyperparameters such as epochs, learning rate, and batch size. The research findings demonstrate the significant accuracy of the sentiment analysis using the BERT algorithm. The first experiment achieves an accuracy of 75%, while the second and third experiments achieve accuracies of 83% each. The results highlight the BERT algorithm's ability to effectively classify user sentiments towards the InDriver application service. This research contributes to understanding user sentiments, providing valuable insights for decision-making and product enhancement. However, the study acknowledges its limitations and suggests areas for further development, including increasing the dataset size or adding additional preprocessing features.
Keywords:	Sentiment Analysis, BERT Algorithm, 7P Marketing Mix, InDriver Application, User Reviews.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	HYBRID PERSONALIZED RECOMMENDATION MODEL FOCUS ON IMPROVED COLLABORATIVE FILTERING
Author:	XU JIAN HUI, MUSTAFA MAN, ILY AMALINA AHMAD SABRI
Abstract:	In recent years, the popularity and acceptance of online education and large-scale open online courses (MOOCs) have significantly increased. The widespread acceptance of the Internet education model has led to the emergence of various educational platforms. With the continuous improvement of online education systems, more and more courses have been added to online education platforms, expanding the group of learners who can benefit from online education. This article aims to improve personalized recommendation algorithms based on collaborative filtering to better meet the needs of users. Firstly, various recommendation technologies and algorithms currently used in recommendation systems were introduced, highlighting their respective advantages and disadvantages. This model can learn the historical behavior of users and generate personalized recommendations based on their interests and preferences. In our research, we found that collaborative filtering technology has high accuracy in recommendation systems, but there are also some limitations, such as sparsity and cold start issues. In order to address these issues, in this study, the Learning Resource Model (LRM) and Learner Model (LM) were combined into collaborative filtering algorithms to create a Hybrid Personalized Recommendation Model (HPRM). Experimental results show that our hybrid recommendation model outperforms traditional collaborative filtering methods in terms of accuracy, recall, and F1. In addition, our model can effectively handle sparsity and cold start issues, thereby improving the performance of the recommendation system. In summary, our research provides an effective hybrid recommendation method for the field of recommendation systems and provides useful references for future research.
Keywords:	Learning Process Optimization, Personalized Recommendation, Cognitive Diagnosis, Collaborative Filtering, Hybrid Personalized Recommendation System.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	J-SELARAS: DEVELOPMENT OF RATIONALISATION ANALYSIS MODEL
Author:	NORISAH ABDUL GHANI, MOHD ADZA ARSHAD, FAIZUL AZWAN ARIFFIN, KAMARUL AZHAR MAHMOOD, MUSTAFA MAN5, MOHD. KAMIR YUSOF, WAN AEZWANI WAN ABU BAKAR, FARAH SHAHRIN
Abstract:	Rate rationalization is a crucial aspect of adjusting tender rates to ensure fair contract amounts, especially in the context of the Malaysian construction industry. This study specifically focuses on rationalizing Bills of Quantity (BQ), a process traditionally done manually with Microsoft Excel templates, which can be time-consuming and affect contract signing timelines. To address this issue, this article introduces the J-Selaras model, a novel approach that utilizes the Z-Score Altman model to refine data and determine reasonable rates. The workflow of the J-Selaras model involves identifying minimum and maximum values, calculating adjusted means, and using a cut-off analysis to identify acceptable rates. Additionally, the model evaluates rates proposed by successful tenderers within predetermined ranges. The provided algorithm outlines the computations and conditions that guide this assessment. Through experimental validation, the effectiveness of the J-Selaras model becomes apparent. In Experiment 1, there is congruence between submitted rates and calculated cut-off values, confirming the model's reliability. Experiment 2 reveals instances where proposed rates deviate from the acceptable range, validating the model's ability to suggest rates based on the cut-off analysis. In conclusion, the J-Selaras model represents a significant advancement in the rationalization process. It aligns submitted rates with reasonable values while adhering to government policies, ultimately expediting contract signing and enhancing efficiency and fairness in the tendering process.
Keywords:	Rationalisation, Bills of Quantity (BQ), Z-Score, Statistic, Model, Digital BQ,
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	A DATA DRIVEN EMBEDDED FEATURE SELECTION AND ENSEMBLE CLASSIFIER FOR HUMAN ACTIVITY RECOGNITION
Author:	S. ANTHONISAMY, P. PRABHU
Abstract:	Human Activity Recognition (HAR) is a complex yet important to assess or monitor humans in the area of healthcare. Itrecognizes activities from a series of observations on the actions of subjects and environmental conditions. HAR research is the basis of many applications including video surveillance, health care, and human-computer interactions because it provides information about the identity of a person, their personality, and psychological state which is difficult to extract. The human ability to recognize another person’s activities is one of the main subjects of study of the scientific areas of computer vision and machine learning. For learning about human activities, HAR can benefit from the usage of Machine Learning Techniques (MLTs) where accuracy is crucial. Individuals have been interested in recording human activities for the past decade but important issues need to be addressed in order to fully utilize technology in human activity information. Though many studies have investigated HARs, there is a need for higher accuracy in classifications and an impending need to find most suitable HAR features. Hence, this work proposes the A Data Driven Embedded Feature Selection and Ensemble (ADDEFE) Machine Learning Model for Human Activity Recognition technique. This research work uses an embedded feature selection and ensemble classifier for recognizing human activity from sensor-based data as these kinds of classifiers have achieved better performances with the use of weighted combinations. The proposed ADDEFE(Grid+SVM) method gives accuracy of 93.4 and 96.4 respectively for WISDN and UCI datasets. Theproposed ADDEFE(Grid+Random forest) method gives accuracy of 91.6 % accuracy for WISDN dataset.Hence the experimental results of the proposed work outperform with traditional methods.
Keywords:	Decision Tree, Embedded Method, Ensemble Classification, Human Activity Recognition, Prediction, Sensors.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	WIRELESS SENSOR NETWORK ROUTING PROTOCOLS, CHALLENGING ISSUES AND PERFORMANCE COMPARISON
Author:	MOHAN KUMAR B N, SHIVASHANKAR, MANJUNATH R, SUMANTH V, MANOHAR KOLI5, NANDEESWAR S B
Abstract:	Nowadays, Nowadays, Wireless Sensor Network (WSN) is viewed as one among the critical advancements in wireless communication. The going with difficulties limits the ongoing conventions for directing in impromptu organizations being explicitly utilized in WSNs. WSN faces the few difficulties because of remote correspondence impacts. In this paper work, the complete outline of sensors utilized in network setting is done for information trade and introduced network testing issues and examination of the directing conventions. The majority of the sensor hubs are detected information from a distance, more appreciated than real sensor hub. In this way, information driven steering methods is to consideration on the transmission of information determined through specific highlights as opposed to information assortment from the sensor hubs. Boundary service is involved to recover remaining energy levels for hubs and furthermore update the energy levels of hubs. To lessen utilization of energy, a legitimate group head determination is required. Concerning WSN, meddling convention is further developed significant flooding algorithms. Another WSN climate takes benefit of sensed data from currently prepared sensor hubs. It is introduced in real pivotal and hurt conditions generally. It also handles node failures, permits for tradeoffs among overhead and delivery metrics during the WSN performance. Use of MATLAB application tools to assess the aftereffects of WSN standard conventions like LEACH, TEEN, APTEEN etc. In the outcomes, we compared the challenging issues of WSN such as energy efficiency, complexity, scalability, delay, robustness, data transmission, sensor location etc. This exploration work will benefit for correspondence analyst and applications in research ground of WSN.
Keywords:	WSN, energy, cluster, protocols, LEACH, TEAN, APTEEN.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	DATA MINING FOR COVID-19 NEW CASES AND DEATH FORECASTING IN INDONESIA
Author:	FREDDY KURNIAWAN SOEBARKAH, LILI AYU WULANDHARI
Abstract:	COVID-19 has been a major threat to Indonesians and the world in many life aspects. Therefore, it is important to predict the covid-19’s new cases and death accurately to anticipate the rise of covid-19 cases. The goal of the research is to develop the time-series prediction model to predict the number of Indonesia's COVID-19 new cases and death. In this research, we conduct to cluster the locations of data before the prediction process. The creation of clustering model is done as the early step before COVID-19 prediction because of so many data variations exists across the 34 provinces. We use K-Means Dynamic Time Warping (DTW) method to cluster the provinces and some comparison in machine learning and deep learning approach for prediction model as much as the number of clusters. The LSTM is chosen as the deep learning approach where we compare between the benchmark from previous research and our proposed model, together with Support Vector Regressor (SVR) as the machine learning approach. Indonesia’s public COVID-19 data with periods from 1 March 2020 to 3 December 2021 is used for training and testing the model. The experiment results show the best number of clusters is three, and from RMSE and MAE, our new cases and new deaths model have lower error and less overfit. The proposed model improved 29.11% higher for RMSE and 42.55% for MAE respectively than the SVR. While it achieves 20.22% improvement for RMSE and 16.24% for MAE respectively than the state-of-the-art LSTM.
Keywords:	COVID-19, Prediction, Time series, Long Short Term Memory, Data Mining
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text

Title:	IMPLEMENTATION OF K-MEANS ALGORITHM USING PHYTON BASED ON PROFITABILITY AND SOLVENCY RATIO
Author:	NELSI WISNA, MAYSHAFIRA LISTIANINGSIH, SYALMA NURINDAH AL-AZHARY, KASTAMAN4, RASWYSHNOE BOING KOTJOPRAYUDI
Abstract:	This research aims to calculate the profitability and solvency of companies and cluster them using the K-Means clustering analysis method. Profitability ratios are retrieved by Return on Assets (ROA) and Return on Equity (ROE). Solvency ratio is calculated based on Debt to Assets Ratio (DAR) and Debt to Equity Ratio (DER). The research was conducted at subcontractor companies registered on the Indonesia Stock Exchange for the 2019 period. The selection sample used purposive sampling method so that 16 companies were found. The research data is secondary data, namely audited financial reports and obtained by through the official website of the Indonesia Stock Exchange. The results of the calculation of ROA profitability show 7 companies in the very good category, while the results of ROE calculations show very good and good categories, namely 11 companies. Solvency calculations for the DAR ratio explain that 16 companies are in the healthy category while the DER ratio shows that 5 companies have a healthy category. Profitability cluster analysis shows that the optimum number of clusters is 3 clusters. Cluster 1 composed of 7 companies, Cluster 2 composed of 1 company and Cluster 3 composed of 8 companies. Solvency cluster analysis shows that the optimum number of clusters is 3 clusters. Cluster 1 composed of 3 companies, Cluster 2 composed of 7 companies and Cluster 3 composed of 6 companies.
Keywords:	K-Means, ROA, ROE, DAR, DER
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2023 -- Vol. 101. No. 19-- 2023
Full Text