		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

July 2023 | Vol. 101 No.13

Title:	UTILIZING TRANSFORMER-BASED DEEP LEARNING FOR INTENT CLASSIFICATION ON TEXT
Author:	RICHARD SIMARMATA, JONATHAN KRISTANTO, ANDRY CHOWANDA
Abstract:	Today, communication is carried out not only between humans but also between machines. Communication between humans and machines is no longer limited to using tools such as buttons or voice commands, but humans can communicate dyadic or two-way communication with computers. This can be done because of the field of Natural Language Processing (NLP). There are several problems exists in building a system that can naturally understand and communicate with human. In communication, there are a number of slang words and local contexts or expressions that might change the meaning of the word. In the communication, there is always an intention of the speaker to the interlocutor. This can improve the machine’s understanding of human words and communicate with humans naturally. This research aims to explore and improve intent classification using deep learning technique. In this study, several models of deep learning based on Transformer are proposed that have the best performance in the field of text intention classification, such as Bidirectional Encoder Representations from Transformer and XLNet. There is also another approach for comparison by combining BERT Embedding with the Long-Short Term Memory model. The data used to train the model is out of scope obtained through a Kaggle website with a label intention of 50. By using metrics such as F1-score, recall, precision, accuracy, and top K accuracy, it can be concluded that the BERT model has good results and is the best compared to other models.
Keywords:	Text Classification, Intent Detection, Deep Learning, Transformer Architecture
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	FORMING COMMANDS FOR VOICE CONTROL IN KALDI ENVIRONMENT BASED ON JSPEECH GRAMMAR FORMAT TECHNOLOGY
Author:	AMER TAHSEEN ABU JASSAR
Abstract:	The work is devoted to the study of the features of voice control of a mobile robot and the processing of voice commands. For these purposes, standard IT technologies are used. An approach to solving such a problem based on various voice recognition systems is considered. Four main speech recognition systems are described. A system for recognizing voice commands for controlling a robot is considered. A distinctive feature of the proposed system is that the system does not require a permanent connection to the Internet, and the autonomous operation of such a system has the advantage of lower requirements for working with it.
Keywords:	Formation, Utility, Management, Voice Commands, IT Technologies
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	PREDICTING STUDENTS’ ACADEMIC PERFORMANCE: DEVELOPING AN INTELLIGENT DATA MINING CLASSIFIER
Author:	OSMAN AHMED ABDALLA MOHAMMED
Abstract:	This study attempts to propose a classification model for predicting students' academic performance by using a decision tree algorithm. The algorithm was applied to relevant attributes such as gender, high school percentage, general aptitude test score, academic achievement test score, grade average point (GPA), absent rate, and other relevant attributes were subjected to the algorithm. According to the study results, the decision tree algorithm outperformed all other classification algorithms in terms of accuracy, precision, recall, and F1-score on a sample of students from Tayma University College, University of Tabuk, Saudi Arabia. The overall accuracy of the obtained algorithm was 89.7%, which means it correctly classified 89.7% of the instances. Precision, recall, and F1-score were also relatively high in both classes. The findings add significantly to the existing literature and demonstrate efficacy.
Keywords:	Data mining, Classification algorithms, Decision tree, Predictive models, Prediction Algorithms
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	STRUCTURE TRANSFORMATION OF INFORMATION ORGANIZATION SUPPORT OF INNOVATIVE PROCESSES AT HIGH-TECH ENTERPRISES ON THE BASIS OF ADAPTATION OF MULTILEVEL METHODS OF SIMULATION MODELING IN THE ENVIRONMENT OF THE DIGITAL ECONOMY
Author:	SERGEY NOVIKOV, ANDREY SAZONOV
Abstract:	The process of innovative development of knowledge-intensive industries should be implemented within a triangle: government, business and education/science, using various end-to-end innovation and digital solutions and project approach necessary to effectively organize the management process. The aim of the research conducted in the article is the qualitative development of methodological tools for the mechanisms of information support of various process groups in the field of innovation, through the use of actual methods of simulation modeling, which include components of machine learning, which as a result will improve the effectiveness of innovative project groups in high-tech enterprises. The scientific hypothesis of the study is the assumption that the effectiveness of various groups of processes associated with information support of high-tech enterprises in the field of innovation, can act as a qualitative basis for a consistent increase in the level of their competitiveness and reliability in the long-term strategic perspective. Scientific novelty consists in theoretical substantiation and development of methodological tools for competent and effective information support of practical activity of the enterprise, within the framework of its innovation policy, with the purpose of subsequent increase in groups of indicators characterizing economic efficiency of innovation and digital projects on the basis of application of simulation modeling methods, including components of machine learning. Theoretical and methodological basis of the study is based on scientific works devoted to the consideration of problems in the field of organization of information support in the field of innovation, scientific developments of profile experts/specialists in the field of artificial intelligence, economic and mathematical modeling. The authors made a multicomponent conceptual model of the organization system of information support for the life cycle of products of innovative type. A dynamic information model has been developed, which will allow the management of high-tech enterprise making optimal management decisions considering the turbulent economic environment. The authors propose an updated mechanism of organizational and economic management of innovative process groups at high-tech enterprises.
Keywords:	Information infrastructure, Kahonen map, Machine learning, Innovation and digital activity, Information assurance, Optimization of management processes, Databases.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	THE EFFECTIVENESS OF ROBUST MIXED EWMA-CUSUM CONTROL CHART ON G-AND-H DISTRIBUTION
Author:	NOR FASHIHAH MOHD NOOR, AYU ABDUL-RAHMAN, ABDU MOHAMMED ALI ATTA
Abstract:	Cumulative sum (CUSUM) chart and exponentially weighted moving average (EWMA) chart are popularly used in statistical process control (SPC) as they can quickly detect small shifts in the process mean. Recently, a Mixed EWMA-CUSUM (MEC) control chart was introduced for better detection of small shifts. Like the EWMA and CUSUM control charts, the MEC chart relies on the normality assumption for optimal performances. In the case that the underlying distribution of the data is non-normal, the chart may no longer be effective in signaling a true out-of-control process. Therefore, this paper proposed the use of a median estimator for Phase II monitoring of location via the MEC chart. The performance of this robust MEC control chart was tested on various g-and-h distributions in terms of the average run length (ARL). It has been found to perform well regardless of the distributional shapes compared to the standard MEC chart which uses the mean as the estimator.
Keywords:	Average Run Length, Mixed EWMA-CUSUM Control Chart, Normality, Robust Estimator, Statistical Process Control,
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	EMAIL SPAM FILTERING TECHNIQUE: CHALLENGES AND SOLUTIONS
Author:	SHARIFAH MD YASIN, IQBAL HADI AZMI
Abstract:	For many years, practically all industries, from business to education, have used electronic mail for either personal or corporate communication. Spam, often known as unsolicited email, can be used to harm any user and computing resource by stealing important data. Email conversations frequently involve sensitive and private information. Emails are therefore useful to scammers since they able to use these facts for bad purposes. The primary goal of the attacker is obtaining personal data through deception email recipients into opening malicious links or downloading attachments. Cyberthreats have grown significantly during the past few years. The most common cybercrime that makes use of spam emails as a tool is phishing. Email phishing has caused significant identity and financial losses. Spam detection and filtering is a critical and important problem. There are numerous strategies that can be utilized to counter email spam. No technique has, however, been shown to be particularly successful. Some approaches, such as applying machine learning, have a very high potential for minimizing the issues with email phishing. Reviewing filtering mechanisms, particularly those used in email, is crucial for understanding how they work and for spotting potential problems. Based on predetermined criteria, a number of papers on spam email were acquired from various digital sources. The most relevant papers that had just been released were chosen. Many researchers are interested in the methods used to filter spam and emails. One of the most significant and well-known methods for identifying and preventing spam is email filtering. These approaches have been contrasted. In order to identify phishing emails, this paper describes a machine learning (ML) approach. It talks about issues and anticipated future developments. In order to categorize phishing emails at various levels of crime, numerous ML models that have been suggested throughout the years are compared and reviewed.
Keywords:	Phishing, Email, Spam, Machine Learning, Analysis of algorithms
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	INFORMATION SERVICES OF USERS OF RUTM UNMANNED TRAFFIC CONTROL SYSTEM
Author:	BUYAN DANZYURYUN, MAXIM KALYAGIN
Abstract:	The priority task of the development of unmanned aviation is to create conditions for flight safety, which is impossible without the use of unmanned aerial vehicle (UAV) flight control system. One of such systems is the Russian unmanned aircraft system traffic management (RUTM), an important part of which is the flight service support system. Based on the analysis of typical actions of users of RUTM system, the article formulates the requirements for their user interfaces and provides a list of necessary services. The issues of organizing the operation of these services based on the database of aeronautical information (AI) intended for storage, processing and provision of up-to-date information to users of the system are also considered. The size of such a base largely depends on the choice of a system of spatial coordinates, with the help of which objects on the map are associated with real objects on the ground. The analysis of existing coordinate systems showed that for the database under consideration, the best option is to store AI in World Geodetic System (WGS) WGS 84 geographic coordinate system. At the same time, the provision to the end user is carried out after recalculating it into the appropriate projection, depending on the requirements for the nature of the distortions. When displaying AI on 3D map, it is proposed to use a cube or an octahedron as a base polyhedron, and a projection of a sphere onto an octahedron as a projection, which makes it possible to halve the redundancy of the initial AI. The novelty of the work is in the fact that in the course of the study, the requirements for user services of the Unmanned Traffic Management (UTM) system were formulated in detail, based on typical user actions, and the size of AI database necessary to ensure the safety of UAV and manned aircraft flights was determined. The formulated requirements can be a useful addition to the regulations adopted by the leading countries, which provide high-level requirements for the necessary user interfaces of UTM system.
Keywords:	Unmanned aerial vehicle (UAV); Russian unmanned aircraft system traffic management (RUTM); airspace; aeronautical information (AI); external pilot (EP); air traffic flow management (ATFM); World Geodetic System (WGS) WGS 84
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	MAIN DIRECTIONS OF TRAINING HIGHLY QUALIFIED STAFF FOR THE OR-GANIZATION OF PRODUCTION AND OPERATION OF UNMANNED AERIAL SYSTEMS
Author:	ALEXEY TIKHONOV, KIRILL SHCHUKIN, VALENTINA FRIDMAN
Abstract:	This article considers the main problems and directions of training highly qualified staff for the organization of production and operation of unmanned aerial systems (UAS) and unmanned aerial vehicles (UAV) on the basis of Moscow Aviation Institute (National Research University) (MAI) of the Ministry of Science and Higher Education of the Russian Federation. The complex of strategic activities of MAI is analyzed, which is structured into 5 main projects, through the implementation of which a stable position at aeromobility market in Russia can be achieved, including the direction of integrated training of highly qualified staff for the global transport markets. As a result of organizational and economic analysis, the conclusion was made that in the current innovative and economic conditions it is necessary to transform the educational process in order to increase the effectiveness of training programs for highly qualified staff for the organization of the development, production and operation of UAS and UAV through the introduction of advanced technologies and forms of learning, such as project-based learning, adaptive learning based on advanced information technology and artificial intelligence, hybrid learning as a symbiosis of learning with face-to-face and virtual seminars and lectures, etc.
Keywords:	Unmanned Aerial Systems And Unmanned Aerial Vehicles; Airomobility Market; Training Of Highly Qualified Staff; Basic And Additional Professional Training Programs; Professional Development Programs; Professional Training And Retraining Programs; Efficiency Of The Educational Process; Digital Human Resources Platform
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	COMPARATIVE STUDY FOR ANDRIOD MOBILE STATIC ANALYSIS ALGORITHMS
Author:	SARA MAHMOUD SHEHATA, ISLAM HEGAZY, EL-SAYED M. EL-HORBATY
Abstract:	Recently, there has been a rapid increase in the use of smartphones, several of which are connected to the internet. Because of the data movement, malware attacks have enormously increased. Malware causes unexpected behavior in smartphones such as strange charges on your phone bill, invasive adverts, contacts receiving strange messages, poor performance, appearance of new applications, abnormal data consumption and noticeable reduction in battery life. Nonetheless, smartphone users remain unprotected from malware attacks. Thus, mobile antivirus applications have been developed to overcome this issue. Since android has established itself as the industry's dominant operating system for smartphones, many antivirus applications are available in the android play store. This paper presents a comparative study of android mobile static analysis. Static analysis is used to classify malware android Apps through meta data file of APK. Furthermore, we used TF-IDF feature extractor and investigate algorithms for static analysis, such as decision tree, naïve bayes, random Forest, K-nearest neighbor, XGB, MLP, support vector machine, logistic regression, adaboost, ,lasso regression, ride regression , ANN and extra trees. We use two datasets small and large “Drebin”. The results of small dataset show that Multi-layer perceptron (MLP) gives the best overall accuracy 98.84% but it takes the biggest execution time around 33.4 seconds and The results of large dataset show that Extra trees gives the best overall accuracy 99.48%.
Keywords:	Mobile Security; Mobile Antivirus; Malware Analysis; Machine Learning; Classification
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	A DECISION-MAKING MODEL FOR QUALITY IMPROVEMENT USING FUZZY LOGIC
Author:	ANASS MORTADA, AZIZ SOULHI
Abstract:	The world has been experiencing a fierce competition on all levels lately, especially in the field of industry, which has pushed manufacturing companies to improve their key performance indicators, especially quality indicators such as yield and rejection rate, in order to better satisfy customers and also to reduce financial losses due to non-compliant products. In order to improve the level of quality and before moving on to solutions, companies aim to first determine the most critical issues in order to prioritize them in the analysis and actions. Among the most used Lean Manufacturing tools in this sense is the Pareto chart which is often used to identify the most critical defects based on a single input indicator which is usually the rejection rate, but this indicator alone is not sufficient to give accurate results to decide which defects are the most critical. The objective of this research is to develop a decision support model using fuzzy logic capable of accurately determining the most critical defects to prioritize based not only on the rejection rate but also on two indicators that have an impact on the criticality of the defects, namely the recovery rate following rework and the rework cost. Then to validate the proposed model, a case study was conducted on the defect data of a car windshield manufacturing plant. The results were compared with those of the Pareto tool, which allowed to reveal its limitations and to retain the proposed fuzzy logic model for the estimation of the criticality of defects in industrial companies.
Keywords:	Quality Improvement, Fuzzy Logic, Artificial Intelligence, Decision-Making, Pareto Chart.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	DESIGN AND PSO BASED OPTIMIZATION OF BILSTM-CNN STACK OF NEURAL NETWORKS FOR ECG SIGNAL CLASSIFICATION
Author:	TATA BALAJI , N.JAYA , G.VENKATA HARI PRASAD
Abstract:	Classifying ECG signals and analyzing the likelihood of cardiac arrest are essential aspects of the medical field. In recent years, artificial intelligence methods like Artificial Neural Networks have been used to create models for ECG data classification. More sophisticated deep learning methods are effective, such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. A CNN was implemented to extract the characteristics from the images to forecast the percentage of porosity in the successive layers of simulated porosity images. Yet formerly, most LSTM-CNN hyperparameters were determined by experience, which frequently did not result in the most outstanding results. The BiLSTM-CNN hyperparameter was modified using the Particle Swarm Optimization (PSO) technique to improve the ability to learn the properties of data sequences. The BiLSTM-CNN-PSO model offers better prediction stability and accuracy. This study showed that the use of CNNs and Bi-LSTM networks can provide accurate classification of ECG signals. The optimized model achieved a categorization accuracy of 99.2%, which is significantly higher than the mean accuracy of 85.3% achieved by the unoptimized model. The results could contribute to long-term cardiac arrest prevention in the study area.
Keywords:	ANN, CNN, Bi-LSTM, hyperparameters and PSO.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	IMPROVING INTEGRITY, SECURITY, AND ACCURACY DURING DEVOPS PROCESS
Author:	HANAN FAHMY , SAMAR SAMIR , MONA NASR
Abstract:	DevOps is a culture that aids in increasing the service delivery speed of an organization. It encourages the development and operations teams for working and coordinates them together to achieve faster and more automated activities. Although DevOps becomes an essential building block in every software life cycle, its processes suffer from data security attacks over time. Therefore, companies and organizations exert considerable effort including investigation time, buying licenses for third-party defending tools, hiring highly skilled security engineers, etc. to avoid these security attacks. Meanwhile, many studies suggest building DevOps frameworks that concentrate on enabling automated processes without giving much focus to the integrity and validation of the data, which might be lost or manipulated due to significant security attacks during the deployment process, the ignorance of handling security assaults costs a lot of damage to the organizations. In the current study, a DevOps framework DevHash is proposed to improve the previous DevOps frameworks by adding new phases detecting and validating common security attacks, such as data manipulation attacks that might occur during the deployment process. DevHash can detect data manipulation attacks much faster than the other previous DevOps frameworks as the proposed framework DevHash includes four main phases: "Coding", "Build Pipeline and Hashing", "Deployment Pipeline and Hashing” and "Health Check". These phases are important in enhancing and validating data security during the deployment process. To assess the performance of the proposed framework DevHash, a comparison has been made between the DevHash framework and a previous traditional DevOps framework; it is applied to six cases including a collection of software companies from various domains with different technologies. Finally, the results showed that the proposed framework DevHash consumes less time than the traditional DevOps framework in detecting major security attacks such as data manipulation attacks, and it helps in increasing the integrity and the accuracy of data as well as the overall reliability of the DevOps processes.
Keywords:	DevOps, Validation, Deployment Process, Security Attacks, Data Manipulation Attacks.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	A TABTRANSFORMER BASED MODEL FOR DETECTING BOTNET-ATTACKS ON INTERNET OF THINGS USING DEEP LEARNING
Author:	ARCHANA KALIDINDI, MAHESH BABU ARRAMA
Abstract:	The Internet-of-Things (IoT) has revolutionized with the increase in data and serves as a link for the ecosystem of various tools, actuators, and sensors, and this is one of the disruptive technologies. IoT enjoys numerous advantages by accelerating communication between various smart objects around us Hence, IoT has become not only an essential part by serving as a medium for growth in advancing technology; the most crucial part is protecting the data from malicious attacks. To resolve Intrusion Attacks ranging from cyber to rule-based, the new advancing AI plays a significant role. Hence, it is well established in the literature that such attacks can be solved by using Machine Learning Algorithms. We analysed various Machine Learning and Deep Learning frameworks for tackling these attacks on the standard dataset N-BaIoT. This extensive analysis has confirmed that the TabTransformer model with SGD, Adam, and Avg & Sub optimizers has exhibited excellent performance, achieving at least 92.33% accuracy. It was observed that deep learning approaches such as Conv-Net or LSTM-based methods have come close to achieving similar results. Furthermore, we have compared the accuracy of our classification method with that of other studies, and the TabTransformer has demonstrated superior performance. This is the first study to conduct a series of experiments on the N-BaIoT dataset using a range of traditional machine learning and deep learning techniques. Our study has achieved state-of-the-art accuracy of 92.330% for the Provision PT-737E device using the TabTransformer model with Adam optimizer.
Keywords:	N-BaIoT; Transformer; Internet of Things; Machine Learning; Deep Learning.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	PREDICTION OF LUNG CANCER LEVELS BASED ON PATIENT LIFESTYLE AND HISTOPATHOLOGICAL IMAGES USING ARTIFICIAL INTELLIGENCE
Author:	SOFYAN EL IDRISSI, IKRAM BEN ABDEL OUAHAB, YASSINE DRIDER, MOHAMMED BOUHORMA, FATIHA EL OUAAI
Abstract:	Histopathology is the fundamental tool used in pathology for more than a century to establish the final diagnosis of bronchopulmonary carcinoma. The phenotypic information on histological images reflects the overall effect of molecular alterations on the behavior of cancer cells and provides a convenient visual readout of disease aggressiveness. However, the human assessment of the histological image can be sometimes subjective and not very reproducible depending on the case. Therefore, computational analysis of histological imaging via artificial intelligence approaches has recently received considerable attention to improve this diagnostic accuracy. Thus, computational analysis of lung cancer images has recently been evaluated for optimization of histological or cytological classification, prognosis prediction or genomic profiling of lung cancer patients. This rapidly growing field is constantly showing great power in the field of medical imaging informatics by producing detection, segmentation or recognition tasks of a very high accuracy. However, there are still several major challenges or issues to be addressed in order to successfully transfer this new approach into clinical routine. In this paper, we use the power of AI to develop an GUI application able to predict lung cancer levels using two techniques: either clinical features based on the patient lifestyle, or by inserting histopathological lung images. The GUI application is friendly to use, fast and average accuracy of 98%. We also did a comparative study of machine learning and deep learning algorithms for lung cancer classification using 2 different databases, then we choose the best ones to be used.
Keywords:	Pathology, Histology, Cytology, Bronchopulmonary Cancer, Artificial Intelligence, Machine Learning.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	A COMPARATIVE ANALYSIS ON THE PERFORMANCE ATTRIBUTES OF SOFTWARE DEVELOPMENT COST MODEL BASED ON WEIBULL LIFETIME DISTRIBUTION
Author:	TAI HEOUN PARK
Abstract:	In this paper, we analyzed the efficiency and cost attributes of the software development cost model applying the Weibull lifetime distribution, which is known to be suitable for the field of reliability because it can express all kinds of various probability distributions. For this study, failure time data detected during normal operation of the software system were utilized, the maximum likelihood estimation was used for the parameter estimation. Conclusively, first, in the result of comparing reference values (MSE, R^2) for efficient model selection as well as reliability attributes analysis using m(t), the Rayleigh and Inverse-exponential models were evaluated as efficient. Second, as a result of analyzing the attributes of development cost, the Rayleigh model showed the best performance. Therefore, as a result of the comprehensive evaluation of the relevant analysis data, the Rayleigh cost model was found to be the best in terms of performance attributes. Through this study, the performance properties of the software development cost model applying the Weibull lifetime distribution were newly analyzed, and the results related to this are expected to be used as primary design data for developers to explore the cost properties in the initial testing process.
Keywords:	Exponential-basic, Inverse-exponential, NHPP, Performance Attributes, Rayleigh, Software Development Cost Model.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	MANAGEMENT OF SECURITIES PURCHASE-SALE WITHIN THE FRAMEWORK OF A DIFFERENTIAL GAME
Author:	MALYUKOV V. , MALYUKOVA I. , KABYLBEKOVA V , BRZHANOV R. , RZAIEVA S. , KHARCHENKO O. , PALAGUTA K. , MIRKO N.
Abstract:	The game model of a continuous process of trading operations with securities (SEC) is considered. It was shown that the controllability of this process can be described from the point of view of the game approach. It allows one to explore and recommend to players a constructive strategy for managing the purchase and sale of securities. The proposed solution is valid for situations of buying and selling securities in the area of both bullish and bearish trends. Moreover, in both situations, the possible worst actions of counterparty players are taken into account. The model proposed in the paper is distinguished by its use of the potential of a bilinear differential quality game with several terminal surfaces, unlike similar solutions in this subject area. In the course of research, an analytical solution has been obtained for a bilinear differential quality game with dependent motions. The article also presents the results of a computational experiment. Theoretical calculations were confirmed by the data of a computational experiment, during which different options for the ratio of parameters that describe the continuous process of trading operations with the securities have been taken into account.
Keywords:	securities, game model, differential game, strategy, buying and selling
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	ASSESSING THE INFLUENCE OF MEMORY-BASED COLLABORATIVE FILTERING METHODS ON CONTEXTUAL SEGMENTS IN MULTICRITERIA RECOMMENDER SYSTEMS
Author:	CHINTA VENKATA MURALI KRISHNA, DAVULURI SUNITHA, BOPANNA VENU GOPAL, A SARVANI, VELAGAPUDI SREENIVAS
Abstract:	Recommender Systems has grown significantly over the last two decades. Memory-based Collaborative Filtering is part of RS and is a powerful technology that has been applied in several well-established commercial applications.However, memory-based collaborative filtering fails to capture the dynamic user opinions in a detailed perceptive since it uses a two-dimensional rating approach.However, multicriteria RS dominates memory-based collaborative filtering with the inclusion of multiple contexts. In addition, significant research has been done to predict user gratification. However, recent multicriteria recommender systemsfail to avoid the significant issues of the curse of dimensionalitydue to the lower number of ratings among multiple dimensions, leading to poor predictions. This paper proposes a new prediction recommender model on multicriteria recommender systems to predict user gratification with the memory-based user and item collaborative filtering approaches used to impute the missing contextsin multicriteria RS. In addition, various regression models were applied to overall and predicted overall ratings. The results indicate that item-item collaborative filtering with Ordinary Least Squares(OLS) regression in multicriteria RS exhibits low Root Mean Squared Error(RMSE), indicating the accurate predictions of user gratification.
Keywords:	Cost Functions, Recommender Systems, Memory-Based Collaborative Filtering, Muti-Criteria Recommender Systems.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	DESIGN OF NFT SMART CONTRACT SYSTEM USING ETHEREUM NETWORK BLOCKHCAIN TECHNOLOGY
Author:	SENDRINO EDGARD, JAROT S. SUROSO
Abstract:	In this digital era, the development of information system and information technology has become an integral part of people's daily lives. As it is today, technology has made human life easier in the rise of variety of new innovations in various fields. One of the most rising technologies is blockchain, this technology is considered as important technology system for businesses, it can affect in performance and security of data. This blockchain has a strong system and strong security such as cryptography, peer-to-peer network technology, smart contracts, and consensus mechanisms. With the rise of blockchain, we know that cryptocurrency has been around it, almost all the people like cryptocurrency as the new digital currency which can affect the economy of a country. Further in coming days, digital currency has been connected to digital arts which is NFT, a new way of calling arts and they have good use case for businesses. In this paper, we will see how to develop and design simple NFT web and minting/buying method using Ethereum testing net/environment.
Keywords:	Blockchain, Smart Contract, NFT, Ethereum, Information System, Information Technology, Web
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	DRIVING MODE SWITCHING MECHANISM BASED ON REINFORCEMENT LEARNING
Author:	YILIN ZHU, SANG-CHUL KIM
Abstract:	Autonomous driving technology can improve traffic efficiency and safety, making the driving process more convenient and comfortable, thus increasing road safety and reducing environmental pollution. In this study, we explore the mode switching problem between automatic and manual driving and propose a driving mode switching mechanism based on deep reinforcement learning. First, we study the background on driving scenes based on safety degree calculation. Then, a real-time online switching mechanism based on deep reinforcement learning is presented. Finally, the proposed algorithm’s effectiveness is verified using a simulation platform. The simulation results show that based on the analysis of lateral position deviation, heading angle deviation, and safety index data, the deep RL method proposed in this project can effectively realize the switch between manual and automatic driving modes. As the results, the mode switching strategy can respond well to the vehicle’s lateral position deviation, heading angle deviation, and longitudinal safety degree and thus improve vehicle operation safety.
Keywords:	Autopilot, Intensive Learning, Actor–Critic Framework
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	CROWD COUNTING AND ANOMALY DETECTION FROM CCTV FOOTAGES USING DEEP LEARNING AUGMENTED WITH CELLULAR AUTOMATA
Author:	DR. POKKULURI KIRAN SREE, DR. G. SRINIVASA RAO, DR. P. SRINIVASA RAJU
Abstract:	Automatic Video surveillance is the need of the hour and interesting research problem to be addressed. We have understood the need of automatic monitoring of the CCTV footages, security is main concern for any country. In the novel work, we have identified people in the footages, counted the number of people in the video and identified the people with abnormal behavior. We have used CNN convolution neural network augmented with cellular automata for identifying people with abnormal actions etc. The developed classifier has achieved 97.6% identifying the people, 91.3% in counting the number of people in a given instance and 78.9% in predicting the people with abnormal actions. The datasets are collected from the Vishnu Society as we have implemented the same in this society.
Keywords:	Deep Learning, Cellular Automata, Video surveillance, Crowd Counting
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	IMPLEMENTATION OF MACHINE LEARNING TECHNIQUES TO DETECT FRAUDULENT CREDIT CARD TRANSACTIONS ON A DESIGNED DATASET
Author:	NAZERKE BAISHOLAN, MUSSA TURDALYULY, SERGIY GNATYUK, KARLYGASH BAISHOLANOVA, KAZILA KUBAYEV
Abstract:	The rise in technology, particularly the increase in online shopping, has made it easier for cybercriminals to obtain and exploit stolen payment card information. Traditional fraud detection systems are finding it increasingly challenging to keep up with the rapid pace of technological advancement, leading to a surge in payment card fraud. Hence, it is essential for companies to continually update their fraud detection methods to keep up with the latest tactics employed by fraudsters. Machine learning algorithms have the ability to analyze large datasets and quickly identify anomalies or deviations from normal behaviour, making them a highly effective tool for payment card fraud detection. By detecting fraud early, organizations can minimize their financial losses and prevent further damage. In this study, we generated a credit card fraud dataset that comprises three types of fraud cases. The dataset is imbalanced, with a ratio of fraudulent transactions at 0.004, making it close to real-world data. To handle the imbalance in the dataset related to credit card fraud detection, we employed popular machine learning models such as Random Forest, Decision Tree, Logistic Regression, and XGBoost. The results showed that XGBoost and Random Forest outperformed the other models on both the training and test sets. However, the Decision Tree algorithm with unlimited depth had the highest average accuracy on the training set and the lowest average accuracy on the test set, indicating that this algorithm should be avoided due to overfitting. In conclusion, our study highlights the significance of using machine learning algorithms for payment card fraud detection. The results demonstrate that XGBoost and Random Forest are the most effective models for detecting credit card fraud in imbalanced datasets. By employing these models, organizations can improve their fraud detection capabilities and minimize the financial impact of payment card fraud.
Keywords:	Fraud Detection, Anomaly Detection, Transaction Fraud Dataset, Imbalanced Dataset, Random Forest, Decision Tree, Logistic Regression.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	A DECISION-MAKING MODEL BASED ON FUZZY LOGIC TO SUPPORT MAINTENANCE STRATEGIES AND IMPROVE PRODUCTION LINES PRODUCTIVITY AND AVAILABILITY
Author:	ANASS MORTADA, AZIZ SOULHI
Abstract:	One of the bases of competitiveness between organizations is their ability to manage and reduce their costs and ensure their economic gains on a permanent basis. In fact, the maintenance and reliability of machines and equipment are two of the aspects that cause the highest costs for companies. For this reason, plant managers have been aiming in recent years to improve their maintenance policies and adopt the most effective strategies while reducing costs. In fact, the adoption of effective maintenance strategies plays a critical role in ensuring the availability and productivity of production lines in order to cover customer orders, ensure their satisfaction, and avoid any type of complaint related to the non-fulfillment of deadlines. Companies often resort to two classical types of maintenance, namely corrective maintenance and preventive maintenance, or even a more general philosophy based on total productive maintenance (TPM). However, these strategies lack the tools and decision-support models to be applied effectively. In this context, this paper aims to develop a decision support model using fuzzy logic to determine the types of maintenance adequate to each type of anomaly in the production lines by estimating the need for preventive action based on the costs of corrective intervention and preventive intervention and also based on the impact of the anomaly on the plant's performance.
Keywords:	Corrective maintenance, Preventive maintenance, Total productive maintenance, Productivity, Availability, Fuzzy logic, Decision-making.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	AN EFFECTIVE IDS FRAMEWORK FOR IOT USING FEATURE SELECTION AND CLASSIFICATION MODEL
Author:	L. SARALADEVE, A. CHANDRASEKAR
Abstract:	Devices and services that are part of the Internet of Things (IoT) bring convenience but are loaded with significant security risks. When protecting the IoT environment, the Intrusion Detection System (IDS) for the network is really important. Thus, a new hybrid IDS model is developed in this research work to classify and detect attacks present in IoT networks. The proposed model integrates feature selection and classification processes implemented using machine learning (ML) models. The Binary-Black Widow Optimization (BBWO) algorithm selects the optimal feature sets from the given datasets, and the Logistic Regression (LR) algorithm performs the binary classification. The proposed model initially performed data preprocessing using min-max scaling normalization for dataset standardization. After preprocessing, the datasets are split into training and test sets for evaluation. Using the datasets, the features are selected optimally using the BBWO algorithm. The classification is performed using the LR algorithm based on the selected optimal feature sets. The performance of this research model is evaluated based on accuracy, detection rate, FPR, f1-score and precision. The results are evaluated individually for both datasets and correlated. The BBWO-LR model obtained 98.83% accuracy, 98.32% detection rate, 99.58% precision, and 98.95% f1-score for the CICIDS-2017 data set. Using the CICIDS-2018 data set, the BBWO-LR model obtained 98.92% accuracy, 98.17% detection rate, 99.76% precision, and 98.97% f1-score.
Keywords:	Internet of Things, IDS, Feature Selection, BBWO, LR, CICIDS-2017, CICIDS-2018.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	ENHANCING USER-LEVEL SECURITY: PERFORMANCE ANALYSIS OF MACHINE LEARNING ALGORITHMS FOR DYNAMIC KEYSTROKE ANALYSIS
Author:	B RADHA KRISHNA, Dr M SRIHARI VARMA
Abstract:	In todays computing and information-processing industry, security has become a paramount concern. With the proliferation of internet-based devices and the exponential growth of data transmission, the threat of malicious activities has also escalated. One prevalent form of attack involves the theft of passwords and unauthorized access to sensitive information, resulting in data loss and diminished user satisfaction. To combat this issue, it is crucial to detect and prevent fraudulent logins. This paper presents a comprehensive performance analysis of machine learning algorithms for dynamic keystroke analysis, aimed at enhancing user-level security. The focus is on developing a robust method that ensures tight security while maintaining high accuracy. Previous research efforts have proposed various security mechanisms, techniques, and algorithms; however, their efficiency has been found lacking. The proposed approach leverages machine learning algorithms, specifically the KNN and XGB models, to analyze dynamic keystroke patterns. By extracting and analyzing important parameters related to keystroke dynamics, the algorithms aim to identify fraudulent login attempts. The performance of both algorithms is evaluated and compared, with a particular emphasis on accuracy and efficiency. The results of the analysis demonstrate that the XGB model outperforms the KNN algorithm in terms of security enhancement and user satisfaction. The XGB model leverages the dataset effectively, utilizing key parameters to make accurate predictions and classify keystroke patterns. In contrast, the KNN algorithm falls short in achieving comparable levels of accuracy and efficiency. By employing the superior XGB model, organizations can enhance user-level security, prevent password theft, and safeguard sensitive data. The findings of this study contribute to the development of an effective and reliable security mechanism, ultimately improving user satisfaction and data protection. In conclusion, this research highlights the importance of adopting advanced machine learning algorithms for dynamic keystroke analysis to enhance user-level security. The XGB model emerges as a powerful tool, surpassing the limitations of the KNN algorithm and providing a more robust and accurate solution. By implementing these findings, organizations can bolster their security measures and address the growing challenges associated with data protection and user satisfaction in the digital age.
Keywords:	Dynamic keystroke analysis, Machine learning algorithms, User-level security, Performance analysis, XGB model, KNN algorithm, Fraudulent logins, Password theft, Data protection
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	ESTIMATION OF MINIMAL INITIAL SAMPLE SIZE IN PROGRESSIVE SAMPLING FOR BIG DATA ANALYTICS
Author:	YATHISH ARADHYA B C, DINESHA H A, LOKESH M R
Abstract:	Big data are vectored, high dimensional and voluminous. Sampling such data is daunting task. Progressive sampling is the solution for such data. However initial sample size of the progressive sampling technique plays a vital role in the overall computational time and convergence time of any classifier. In Progressive Sampling Algorithm (PSA) a number of times iterative computation runs will be accountable for the total time cost of the sampling and indirectly to the time required for convergence of the classifier to learn a hypothesis. All existing works on minimal sample size estimation are not appropriate to carry out in the Distributed File system like Hadoop. In this work we present a novel statistically optimal sample size technique and its analysis, to estimate the initial minimal sample size for big data in an HDFS environment. Heterogeneous big data datasets were experimentally used to estimate initial sample size in a Hadoop environment with the analysis of computational time and space complexity in all degrees of freedom along with the convergence of the learning algorithm. If the initial sample size were accurately estimated, then there will be a substantial reduction in PSA. Thus providing a proper initial sample size for PSA will ensure optimally fast learning of the classifier in Information Technology applications for substantial prediction and assessments thus leading to robust software performance.
Keywords:	Progressive Sampling Algorithm (PSA), PAC Framework, Big Data, Sample Size, Initial Sample Size
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	A NOVEL LOW COMPLEXITY ALGORITHM FOR OFFENCE GESTURE DETECTION IN LIVE VIDEO STREAM
Author:	ABID SIDDIQUE, MOHAMMED GHOUSUDDIN, REHNA V. J.
Abstract:	One of the major causes of road accidents is the use of mobile phones while driving. Calling or texting while driving, risking the life of the driver, passengers and other road users, is considered to be a traffic offence in most countries. The stringent measures taken by the authorities including increased penalties and jail terms, has made little effect, especially on the younger sections of the driving community, due to their addition to smartphone usage. The aim of this study is to provide a possible solution for this concern. A novel efficient gesture recognition algorithm for sensing specific gestures in real time while driving is presented in this paper. The offensive actions such as talking over cell phone or texting while driving are identified from real time live video streams captured in a non-smart environment using image processing algorithms, and the data is used to alert the user to refrain himself from improper driving conditions and/or send to concerned authorities for storage and subsequent action. The algorithm was tested using MATLAB 9.3 R2017b on over 60 sample videos. Experimental results demonstrate that the proposed algorithm is competent with successful detection rates approaching 98.3%. The time complexity of the algorithm is 1.5 times lower than its contemporaries and so is its computational complexity.
Keywords:	Image Processing, Texting-And-Driving, Offence Detection, Gesture Identification
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	CLASSIFICATION ANALYSIS FOR LAND SUITABILITY USING LINKED OPEN DATA
Author:	CHARITAS F., AHMAD A., MARDHANI R.
Abstract:	Land suitability is one of a solution to get alternative solutions to get maximum results. Land suitability is obtained by applying classification rules based on several factors, such as: nutrients, erosion hazard, temperature, flood hazard, and root media. The analysis will classify a land based on its order class into 2, such as: suitable (S) and non-suitable (N). Spatial analysis for land suitability usually put together all the required spatial data into one source first, and then analyzes it using land evaluation rules. However, the concept of linked open data can create structure that are connected between data from different sources, including applying classification rules to these data. Information related to the required attributes can be read using LOD concept. The formulation of the problem in this study is how to classify the suitability of a location for rice plants, if the data to be used as measuring variables are at different storage sources. This research used a coordinate of an area as an identity that is used as a linked between different sources. In addition, it will obtain the information that is needed for land suitability then classification rules are applied based on information obtained from that location.
Keywords:	Linked Open Data, Spatial Analysis, Information Intelligent, Precision Agriculture, Land Suitability
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	GAME THEORY BASED EFFICIENT ROUTING PROTOCOL IN MOBILE WIRELESS SENSOR NETWORK USING ADVANCED LEARNING MECHANISM
Author:	KAMALINDER KAUR, SANDEEP KANG
Abstract:	Mobile Wireless Sensor Network (MWSN) plays a paramount role in controlling, monitoring, and collecting the required information from the sensors. MWSN consumes less energy and controls the lifespan of the network. The main challenge of using the MWSNs is the use of a routing mechanism, aimed to transfer the sensor information to the sink. In this paper, a novel Game theory (GT) based Routing Protocol (RP) in an MWSN has been developed. The proposed protocol is abbreviated as GT-RP in which fault tolerance has been achieved by offering alternative routes to pass the data when the existing route includes the presence of a fault. The game theory model has been improved further to route the data and then Machine Learning based Q-learning technique has been applied for the selection of alternative routes to maximize the data throughput rate and PDR with minimum delay. The improvement shown by the simulation results is (~ 5.3%)for the PDR, (~ 13.24%) for throughput and (~ 6.48%) for delay against existing techniques.
Keywords:	MWSN, Mobility, Game Theory, Machine Learning, Routing Protocol.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	WHAT IS THE MISSING? COGNITIVE DISTORTIONS AND ADOLESCENT CYBER DELINQUENCY BEHAVIOUR: PLS-SEM ANALYSIS
Author:	FAHAD NEDA ALENEZI, SHAHABUDDIN BIN HASHIM, JAMALSAFRI BIN SAIBON, MASHAIL ALANEZY
Abstract:	Cyber delinquency among school adolescents has become a significant concern associated with negative classroom social climates. However, understanding the cognitive distortions contributing to their involvement in electronic delinquency remains limited. While previous research has linked cognitive distortions to anti-social behaviour, such as delinquency, few studies have explicitly examined this association in the context of cyber delinquency among school-delinquent adolescents. Drawing on Beck's cognitive theory, this study aims to fill this gap by investigating the relationships between specific self-serving cognitive distortions (blaming others, self-centeredness, minimising/mislabeling, assuming the worst) and cyber delinquency behaviour. The measurement of cyber delinquency includes three dimensions: cyberbullying, harassment, and impersonation. The How I Think Questionnaire (HIT) is used to assess cognitive distortions. The data was collected from 386 adolescent delinquents aged 18, in grade 12, from secondary schools in Saudi Arabia. The study utilises PLS-SEM as an advanced statistical approach to test the hypotheses. The model demonstrated a proper fit and revealed significant positive relationships between self-centeredness, blaming others, assuming the worst, and cyber delinquency behaviours. These findings contribute to enhancing the understanding of the underlying mechanisms that drive engagement in cyber delinquency among school adolescents. By investigating specific cognitive distortions, this research sheds light on adolescents' inclination to disengage and provides insights into the occurrence of delinquency behaviours in the online environment.
Keywords:	Cyber delinquency
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	EFFICIENT BIG DATA SECURITY: EVALUATING THE PERFORMANCE OF A PROPOSED HYBRID KEY MANAGEMENT ALGORITHM USING LIGHTWEIGHT CRYPTOGRAPHY
Author:	MARWA KHADJI, SAMIRA KHOULJI, MOHAMED LARBI KERKEB
Abstract:	This research paper explores the integration of lightweight cryptography algorithms into the MapReduce framework for secure and efficient big data processing. Initially, lightweight cryptography was not a priority in the field of big data due to the focus on scalability, security trade-offs, potential performance impact, and compatibility with existing solutions. However, as the need for lightweight and efficient security measures emerged, researchers began to explore the integration of lightweight cryptography into MapReduce or develop specialized lightweight cryptographic solutions for big data processing. To address the resource-intensive nature of traditional cryptographic algorithms in big data applications, this paper proposes a new algorithm with a hybrid key management scheme that leverages the efficiency and effectiveness of popular lightweight cryptography algorithms. The proposed algorithm ensures secure data processing while maintaining low computational overhead. Experimental evaluations were conducted using different big data scenarios to measure processing time, memory utilization, and security of the algorithms. The results demonstrate that lightweight cryptography algorithms, such as Rabbit stream cipher and NOEKEON block cipher, offer practical and efficient solutions for securing large volumes of data. Depending on the specific requirements, AES and Chacha20 algorithms can be selected for confidentiality and integrity. Additionally, Rabbit stream cipher and NOEKEON block cipher are the preferred options when high speed is a crucial factor. These findings provide insights into the practical implementation of lightweight cryptography in MapReduce for secure and efficient big data processing.
Keywords:	Big Data, Hadoop, Data Security, MapReduce, Lightweight Cryptography Algorithms.
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	SENTIMENT ANALYSIS ON INVESTMENT TOPIC IN INDONESIA USING MACHINE LEARNING ALGORITHMS APPROACH
Author:	FLAVEGA MONESA, RIYANTO JAYADI
Abstract:	Investment is one of the rising topics in Indonesia after the pandemic. To be able to understand the overall sentiment or public opinion towards the investment topic, it is important to monitor its growth of it through opinions on social media. In this study, an automated sentiment analysis on the Investment topic in Indonesia is conducted. The proposed technique automatically labeled the sentiment as positive and negative. The opinion data used is in Bahasa Indonesia. To train the model, opinions were scrapped from tweets through Twitter Developer API. There are 10,000 tweets in Indonesian scrapped from September to November 2022 using “Investasi” keyword This analysis uses Naïve Bayes Classifier and Support Vector Machine model to compare which algorithms has better accuracy. The result shows that the SVM algorithm works better with an accuracy rate of 95,7% compare to Naïve Bayes Classifier with an accuracy rate of 94,6%.
Keywords:	NLP, Sentiment Analysis, Machine Learning, Naïve Bayes, Support Vector Machine (SVM)
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	DESIGN OF GAME-BASED LEARNING DEFEND THE COUNTRY USING ROLE PLAYING GAMES MECHANISM
Author:	RISKA NURTANTYO SARBINI, IRDAM AHMAD, ROMIE OKTOVIANUS BURA, LUHUT SIMBOLON
Abstract:	Nowadays, it is undeniable that smartphones have been widely used by all levels of community, including children. Besides being used for communication media, smartphone is also used as a media for digital games. There are various forms of digital games, one of which is game based learning. Learning using video game media is an alternative learning method which has been applied at outside school, but it has not been accommodated by formal education. In our research, we try to combine the learning process with video game games in studying defend the country with the type of Role-Playing Games (RPG). Based on the calculating results from initial conditions of players, it was obtained a value of 75.88% in the pretest phase and there was an increase with a value of 12.88% to 88.75% in the middle test through traditional learning methods, and had increased by 17% to 92.88% at the end of evaluation by game learning methods. Furthermore, analysis towards the performance of this game obtained a total average value of 87% from five aspects that had been tested. In addition, it would facilitate students to study independently, thus maximizing the time available for elementary school students to study and increase character values for being better in the future.
Keywords:	Game Application, Learning Media, Game-Based Learning, Role Playing Games, Defend the Country
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text

Title:	VOWEL RECOGNITION FOR SPEECH DISORDER PATIENT VIA ANALYSIS ON MEL-FREQUENCY CEPSTRAL COEFFICIENT (MFCC) IMAGES
Author:	NUR SYAHMINA AHMAD AZHAR, NIK MOHD ZARIFIE HASHIM, MASRULLIZAM BIN MAT IBRAHIM, MAHMUD DWI SULISTIYO
Abstract:	An individual with a speech disorder, autism, brain injury, autistic spectrum disorders, and stroke usually has trouble producing or forming the spoken sounds necessary for effective interactions. As a result, patients' rehabilitation and treatment typically take a long time and involve ongoing medication, physical activity, and rehabilitation training. However, this rehabilitation process is still done manually in most rehab centers worldwide. Since the impact of computer vision on this profession, machine learning and deep learning have been introduced to the medical industry to improve rehabilitation using the new technology. Convolutional Neural Network (CNN) models have been proven in countless studies to be precise at classifying performance in various fields, including visual field, computer vision, audio, and text defects. This study analyzed the classification accuracy of different pre-trained models (Designed network, VGG-Net, AlexNet, and Inception). We created a thorough comparative analysis to compare the accuracy of several CNN models. The image-profiled sound uses the Mel-frequency Cepstral Coefficient (MFCC) to produce the best results and accuracy. This study aims to create a neural network that can discriminate between the vowels in the voices of normal persons and speech disorder patients. According to experimental results, the designed model had the highest accuracy of 94.54% by using 6 batch sizes, 20 epochs, and ADAM as the optimizer. Furthermore, we discovered that combining various hyper-parameters and fine-tuning the pre-trained models deliver impacts the performance of deep learning models for this classification task.
Keywords:	Convolutional Neural Network (CNN), Deep Learning, Mel-Frequency Cepstral Coefficient (MFCC), Speech Disorder Patients, Vowel Recognition
Source:	Journal of Theoretical and Applied Information Technology 15^th July 2023 -- Vol. 101. No. 13-- 2023
Full Text