		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

November 2023 | Vol. 101 No.21

Title:	TRANSFORMATION OF URBAN CITIES TO SUSTAINABLE SMART CITIES - CHALLENGES AND OPPORTUNITIES FACED BY SAUDI ARABIA
Author:	DR. MURAD ANDEJANY, DR. ARIF MALIK, DR. WAQAR AHMAD, DR. ABDULLAH M. ALHARBI, DR. SYED UMAR
Abstract:	Transformation of urban cities to sustainable smart cities is growing rapidly in last few decades. This process has strong impact on country’s infrastructural and financial growth. This paper focused on the overall challenges and opportunities in transformation of urban cities to sustainable smart cities in Kingdom of Saudi Arabia. Key study objectives are to show the significance of transformation trend mapping with urban influences and to address the challenges and opportunities of sustainable smart city transformation process. Literature has shown significant evidence of taking further initiatives to bridge the gap in the process of urban cities sustainable transformation. Study has shown the importance of transformation of urban cities to sustainable smart cities based on the literature and discussed the challenges and opportunities as per the global standards recommended by UN in Sustainable Development Objectives (SDGs). For this purpose, first study details the population of Saudi Arabia, in terms of its growth, based on gender and age group and showed the population stats of Saudis and Non-Saudis living in the Kingdom, until the mid-year of 2021. Secondly, study further observed the overall impact of transformation process aligned with the seventeen UN-SDGs by showing a trend mapping of transformation process of sustainable smart cities. Study has shown a comparison of local and global advancements in transformation of urban cities. Study has also pointed out some challenges like increase of capital cost, lack of skilled labor, privacy and security concerns, policy based framework, establishing the competitive environment, public dependency on public transport and internet related security and connectivity risks needs. Study also focused on opportunities like automation, efficient administration, eco-friendly sustainable environment, analytical use of data and better connectivity with efficiency can also be a great opportunity for a sustainable smart city transformation process.
Keywords:	ICT, IOT, Smart Cities, Sustainable Development, UN-SDGs.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	DEVELOPMENT OF DECISION SUPPORT SYSTEM IN DETERMINING PROSPECTIVE STUDENT RECIPIENTS OF THE PROGRAM INDONESIA PINTAR USING THE WEIGHTED PRODUCT METHOD
Author:	MELDA AGNES MANUHUTU, ROXIMELSEN SURIPATTY, JALIMJN TINDAGE, TAGOR MANURUNG, MARISSA TUPAMAHU, LULU JOLA UKTOLSEJA
Abstract:	Program Indonesia Pintar (PIP) is a national program that provides educational cash assistance to all school-aged children (6-21 years) who receive the Kartu Indonesia Pintar, or who come from poor and vulnerable families (for example, from families or households holding the Kartu Keluarga Sejahtera or children who meet predetermined criteria. PIP aims to remove barriers for students to attend school by helping poor students gain access to more appropriate education services, preventing children from dropping out of school, helping underprivileged children meet their needs in school activities, and supporting the completion of the 9-year compulsory basic education and universal secondary education.PIP distribution has been carried out by the government in almost all schools in Indonesia. SMP Negeri Persiapan Supnin is one of the recipients of the PIP scholarship. PIP is a scholarship that can be provided by the government for students of SMPN Persiapan Supnin which are categorized as incapacitated with reference to several criteria, such as student active status, certificate of inability, orphan condition, parents' salary, and absenteeism percentage. The PIP distribution process caried out at SMP Negeri Persiapan Supnin has not been optimal so far. The researcher design a study entitled Development of a Decision Support System in Determining Prospective Student Recipients of the Program PIP at SMP Negeri Persiapan Supnin in Raja Ampat Regency Using the Weighted Product Method. Decision support system technology is used as an objective decision making solution because the use of information systems will minimize subjective judgments in determining prospective scholarship recipients.
Keywords:	Decision support system, Weighted Product, Student, Program Indonesia Pintar
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	PERFORMANCE ANALYSIS OF INTRUSION DETECTION APPROACHES IN AGRICULTURAL IOT NETWORKS: A MACHINE LEARNING BASED APPROACH
Author:	DR. S.B. DEOSARKAR, SANTOSH KONDE
Abstract:	In the last decade there is a tremendous growth in the field of IOT ,which is related with Agriculture. The continuous development in the Agriculture IOT, there is number of connected devices have large amount complex network data and huge amount of complexity. In the Agriculture IOT Network have connected the low power CPU and minimum memory capacity, Live streaming devices. These devices are critical decision dependent which are unable to run security software in the existing environment. These Agriculture IOT devices create the inherent risk in the Networks. The attackers have more concentrated on this Agriculture IOT devices. The number of possibility on the Network attacks are increasing due to these draw backs. The current intrusion detection system (IDS) is not working effectively. For the analyzing and dealing with improvement study in the field of IDS and prevention techniques will be identify to the normal and abnormal activities in the field of IOT Agriculture Networks. Thus there is a requirement to design the effective IDS using machine learning for the field of Agriculture IOT Networks. In this paper we have represent the survey and comparative study with the analysis of the machine learning methods, to threat detection in the field of Agricultural IOT networks environment. Machine learning methods performed on NSL – KDD and BOT – IOT data set. In this method we analyze machine learning (ML) models that include support vector machine (SVM), decision tree(DT), Naïve Bays(NV), Random forest(RF), and K – nearest Neighbors (KNN). Also we used the most performance Indicators namely as accuracy, precision, recall and F1 score, to test the effectiveness of several methods.
Keywords:	Internet of things (IOT), Machine learning (ML), Intrusion detection system (IDS), Dataset, Algorithm.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	MASKED FACE DETECTION: ADVANCEMENTS, CHALLENGES, AND RESEARCH DIRECTIONS
Author:	SANTO CHERIAN, BINU THOMAS
Abstract:	Facial identification has become indispensable in our regular lifestyle as a convenient and quick method of precisely validating identity. It has improved and expanded in lockstep with technological advances and intense learning. The most recent COVID-19 epidemic has highlighted the significance of hygienic and contactless identity validation. Individuals have been obliged to wear masks in order to limit the transmission of the coronavirus. However, this makes it impossible to monitor sizable crowds of mask-wearing people. The effect of using a mask on group facial identification can be a contentious subject that has yet to receive much attention. The paper overviews the various AI algorithms employed for masked face detection and their associated datasets. Current benchmarking initiatives that primarily focus on masked face identification algorithms are reviewed. Furthermore, this paper examines existing evaluation efforts that primarily concentrate on assessing the performance of algorithms specifically designed for identifying masked faces. The identified research directives could be an excellent starting field for researchers looking to create increasingly efficient and productive systems in masked face detection
Keywords:	Masked Facial Detection, Neural Network, Datasets, Face mask, Deep Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	A MULTI-EXPERT APPROACH TO IMPROVING DICTIONARY-BASED SENTIMENT ANALYSIS WITH CORRELATED FOREX MARKET DATA
Author:	ALEXANDER MUSAEV, DMITRY GRIGORIEV
Abstract:	One challenge in automatically extracting knowledge from text documents is processing messages where the content is presented implicitly or in a veiled manner. Traditional text analyzers, which rely on search methods for analyzing pre-selected content features, may not provide the desired conclusions. To address this issue, it is proposed to analyze the tonality of the text using an expanded dictionary of search features that covers a segment of correlated concepts. This approach was applied to speculative management of financial assets in day trading on the Forex market.
Keywords:	Multidimensional Non-Stationary Processes, Multi-Expert Control, Distributed Decision, Multi-Expert Asset Management, Knowledge Extraction.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	INTEROPERABILITY AND EXPLAINABILITY OF MACHINE LEARNING CLASSIFIERS TO DETECT LUNG CANCER
Author:	JYOTIRMAY DEVNATH , MD. NAHID SULTAN, MD. FERDOUS WAHID, AHSAN HABIB
Abstract:	The prominence of lung disease as the leading cause of death in cancer necessitates utmost significance on early detection, prediction, and diagnosis of lung cancer, owing to time limitations and the intricacies of the ensuing clinical examination. Hence, the use of machine learning (ML) models may enable the early stage diagnosis of cancer as well as the characterization, stratification, and consequences of the disease. Therefore, several machine learning algorithms have been used in this paper to predict lung cancer, including Logistic Regression (LR), Support Vector Classifier (SVC), Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB), Multinomial Nave Bayes (MNB), Gradient Boosting Classifier (GBC), k-Nearest Neighbor (KNN), and Adaptive Boosting classifier (ABC). To make the data more trainable for the ML models, we proposed a preparation pipeline that included data cleaning, normalization, and data balancing. Nevertheless, healthcare practitioners may exhibit reluctance in embracing artificial intelligence (AI) models if the reasoning behind the generated predictions remains inscrutable. As a consequence, explainable artificial intelligence (XAI) is gaining popularity to meet the needs of healthcare practitioners. Hence, we employ XAI tools (SHAP and Shapash) to rank features, find partial dependencies, and correlate top feature dependencies to find the inner pattern of the features. We investigate both the global and local explainability of the ML model. RF, LR, and XGB among all algorithms exhibit 95% accuracy. To demonstrate the reasoning behind the prediction, we use XAI tools on RF. When it comes to lung disease, Allergy, Coughing, and Swallowing difficulty are of utmost importance. The expertise of the domain expert might be mapped to the developing field of XAI using this research.
Keywords:	Lung Cancer Prediction, Machine Learning, Classification, Explainable AI, Model Interpretability
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	A NOVEL INTRUSION DETECTION SYSTEM (IDS) FRAMEWORK FOR AGRICULTURAL IOT NETWORKS
Author:	SANTOSH KONDE, Dr. S.B. DEOSARKAR
Abstract:	The Tremendous growth of the Agricultural Internet of Things (IOT) applications has required a huge amount of network data and created high computational complexity across various connected devices. IOT devices capture valuable information that enables users to make critical decisions dependent on live streaming. Most of these Agricultural IOT devices have resource limitations such as low CPU, limited memory, and low energy storage. Thus, these devices are vulnerable to attacks due to the lack of capacity to run existing security software. This creates an inherent risk in Agricultural IOT networks. This has resulted in attackers having more incentive to target IOT devices. If the hackers attacking on the networks; the traditional intrusion detection system (IDS) cannot detect threats effectively. Therefore, there is a need to develop effective IDS using machine learning (ML) techniques in the Agricultural IOT networks. In this paper, we propose IDS, which is a combination of feature selection and classification. We suggest using Pearson's correlation coefficient based feature selection and K-Nearest Number (KNN) based classification method to detect attacks in an Agricultural IOT networks. This will increase the accuracy of the classification and reduce the complexity of the system by extracting only nineteen key features from the original Forty One features in the dataset. The performance assessment of the proposed IDS was conducted using tests conducted on the intrusion benchmark dataset NSL-KDD. In this work, we compared the proposed IDS with other ML models including Support Vector Machine (SVM), Decision Tree, Naive Bayes, Random Forests and KNN. Additionally, we used the most important performance indicators, namely, accuracy, precision, recall and F1 score, to test the effectiveness of proposed IDS. The results obtained show that our proposed IDS can effectively reduce the number of features with higher classification accuracy compared to other ML-based classification methods.
Keywords:	Internet of things (IOT), Machine learning (ML), Intrusion detection system (IDS), Dataset, Algorithm, framework.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	IMPLEMENTATION OF INCREMENTAL APPROACH IN R-DIFFSET FOR INFREQUENT ITEMSET MINING
Author:	JULAILY AIDA JUSOH, SHARIFAH ZULAIKHA TENGKU HASSAN, WAN AEZWANI ABU BAKAR, MOHD KHALID AWANG, SYARILLA AHMAD SAANY, NORLINA UDIN @ KAMARUDDIN
Abstract:	Data mining is a well-established approach for extracting crucial information from databases that employs the Association Rule Mining (ARM) technique. It can unearth hidden information that can help with decision-making, financial forecasting, marketing policy, medical diagnostics, and other uses. ARM is the most widely used data mining approach for discovering exciting correlations and connection pattern among itemsets in transaction databases. This vital data can lead to the association rule, suggesting a positive trend. The advantageous itemset of the association's regulations is typically expressed as frequent and infrequent. There are two data formats in itemset mining that is horizontal and vertical. R-Eclat, or Rare Incremental Equivalence Class Transformation, is an example of a vertical data mining approach for an infrequent itemset. The R-Diffset variant, one of four R-Eclat algorithm variants, will be the focus of this study. Previous research has shown that the R-Diffset algorithm takes a long time to process data. Current research outcomes in infrequent mining techniques focus on vertical data formats. The experimental result this indicates that the comparison analysis for three (3) datasets that is mushroom, pumsb_star, and chess. The average performance in terms of the execution time of IR-Diffset is better than R-Diffset.
Keywords:	Data mining, Asscosiation Rule Mining (ARM), Infrequent itemset mining, R-Eclat algorithm, R-Diffset algorithm, IR-Diffset algorithm
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	MAPPING THE PHISHING ATTACKS RESEARCH LANDSCAPE: A BIBLIOMETRIC ANALYSIS AND TAXONOMY
Author:	MELTEM MUTLUTÜRK, BILGIN METIN
Abstract:	Phishing attacks represent a worldwide issue that requires a comprehensive, global strategy to tackle. Delving into international scholarly research on phishing incidents allows us to grasp the extent of this problem on a global scale, while taking into account the unique obstacles and perspectives that arise in different regions. This bibliometric study offers a comprehensive analysis of the phishing research domain from 2004 to 2023, highlighting the growth, trends, and collaborative networks shaping this field. The presented study uncovers the most influential articles, authors, and institutions, as well as emerging research themes and collaborative patterns using network analyses, including citation, co-citation, co-authorship, co-occurrence, and bibliographic coupling. The results demonstrate a consistent growth in the number of publications, indicating the increased interest and relevance of phishing research in addressing cybersecurity challenges. The study identifies the main research clusters and emerging topics, offering insights into future research directions and practical applications. Furthermore, the analysis emphasizes the importance of fostering interdisciplinary collaboration and academia-industry-government partnerships to develop more effective countermeasures against phishing attacks. By understanding the current research landscape and promoting stronger partnerships, stakeholders can work together to devise innovative strategies and tools to protect individuals and organizations from phishing threats. Lastly, the study provides a taxonomy of the phishing literature.
Keywords:	Phishing, Bibliometric, Taxonomy, Vosviewer, Collaboration
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	ENHANCING AGILE DEVELOPMENT WITH SECURITY INTEGRATION: INTRODUCING THE HSSCRUM FRAMEWORK FOR OPTIMIZED AND SECURE SOFTWARE DEVELOPMENT
Author:	SYED SHABBEER AHMAD, AMOGH DESHMUKH, MASRATH SABA, IMTIYAZ KHAN4, D.SHRAVANI, M.UPENDRA KUMAR
Abstract:	We proposed an agile software process model with security integration. It is known as “HSScrum” framework that is an integration of traditional Scrum model being used widely as process model for software development and a security process which is hybrid and flexible to leverage productivity and optimize the development process. HSScrum has Scrum based functions with security provisioning and a novel security process that is seamlessly integrated with the Scrum model to realize HSScrum. As the traditional Scrum with security provisioning contains necessary phases in the System Development Life Cycle (SDLC), it needs integration of security process that is more beneficial and ensures that the development process is optimized. HSScrum realizes this objective with the loosely coupled (in the sense of flexibility) security process integrated with Scrum with security provisioning. HSScrum has a risk identification process that not only finds risk and rank the user stories based on risk, it also has provision to know whether risks are specific to a backlog item or multiple backlog items (cross-cutting security concern). The mapping and delegation process has mapping of security concerns to backlog items and also a hybrid approach in delegation is preferred. Based on the security expert availability and cost analysis, the delegation may be immediate delegation or deferred delegation. This brings about balance between cost and faster intermediate deliverables to client. Our empirical study has revealed its faster convergence and security capability.
Keywords:	Agile Process Model, Software Engineering, Security Framework, Agile With Security, Software Process Mode
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	CLASSIFICATION OF FUNGAL SPECIES USING K-NN BASED ON COLOR FEATURE EXTRACTION AND GLCM
Author:	WILIS KASWIDJANTI, BAMBANG YUWONO, INDAH WIDOWATI, DILA AJENG MEILIAWATI, MANGARAS YANU FLORESTIYANTO
Abstract:	In digital image processing, feature extraction is an important task to obtain crucial information about the characteristics of the image. One of the feature extractions that can be analyzed is texture feature extraction. Grey-level Co-occurrence Matrix (GLCM) is a texture feature extraction method that uses statistical approaches and has been proven to be the strongest descriptor for data classification. Many parameters in GLCM can be used as texture feature extraction values, but some parameters are often used in research, namely ASM, Contrast, IDM, and Correlation. This study will combine several commonly used GLCM parameters with Energy and Dissimilarity parameters. The combination of these GLCM parameters will be implemented to classify the types of portobello and shiitake mushrooms K-Nearest Neighbor (K-NN) algorithm as a classification method. Based on several models that have been constructed, the best performance is achieved when the model is built using 6 parameters, which results in an accuracy of 97%. This accuracy was obtained by testing the system using a confusion matrix with several experiments based on the constructed model and the predetermined K value.
Keywords:	Grey level Co-occurrence Matrix, K-Nearest Neighbor, Fungal, Confusion Matrix
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	CURRENT CHALLENGES AND FUTURE DIRECTIONS IN ARTIFICIAL INTELLIGENCE FOR IMAGING INFORMATICS
Author:	J. RAVINDRA BABU, BHARGAVI PEDDI REDDY, VANGIPURAM SESHA SRINIVAS, A. L. SREENIVASULU, K V S S RAMAKRISHNA, D N V SATYANARAYANA, C.D.VARAPRASAD
Abstract:	Artificial Intelligence (AI) has made significant strides in healthcare, revolutionizing various aspects of medical diagnosis, treatment, and patient care. However, the adoption of AI in medicine is hindered by challenges related to model interpretability, generalization across different healthcare domains, and data privacy concerns. This research paper explores the concepts of explainable AI (XAI), domain adaptation, and federated learning in the context of healthcare, and their potential to address these challenges. Discusses the significance of developing AI models that are explainable, adaptable across different domains, and capable of leveraging distributed data sources through federated learning to enhance medical decision-making while maintaining patient privacy.
Keywords:	Artificial Intelligence, Explainable AI, Federated Learning, Medical.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	EARLY PREDICTION OF GESTATIONAL DIABETES USING MACHINE LEARNING TECHNIQUES
Author:	MOHAMMED ALOTAIBI, NOJOOD ALJEHANE
Abstract:	Gestational diabetes is a form of hyperglycemia that manifests itself in pregnant women. It's possible to experience complications during and after giving delivery if this happens to you at any point in your pregnancy. Particularly in locations where only occasional examinations of pregnant women are available, the hazards can be decreased if they are discovered early and handled. The healthcare industry is not immune to the widespread transformation brought about by intelligent systems developed using machine learning algorithms. This research suggests a combined prediction model for identifying pregnant women who may develop diabetes. The dataset was obtained from the Kaggle, Gestational Diabetes Mellitus (GDM DataSet), which includes records of 3526 pregnant women. Eight models including traditional (Support Vector Machine, Naive Bayes, Random Forest, Logistic Regression, XGBOOST, Decision Tree, SGD) and deep learning (Artificial neural nets) models were used, and the findings resulted in an accuracy ranging from 87%-97% across the models. The results show that deep learning models can significantly improve prediction accuracy.
Keywords:	Gestational diabetes, Machine learning, Medical technology , AI, Diabetes; pregnancy
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	THESAURUS-BASED QUERY EXPANSION ON INFORMATION RETRIEVAL TO IMPROVE THE QUALITY OF DOCUMENT SEARCHING RESULT
Author:	YIYI SUPENDI, ERWIN YULIANTO, DESHINTA ARROVA DEWI, KM SYARIF HARYANA
Abstract:	With the role of the internet as an unlimited source of information originating from various people around the world with a variety of languages and various forms of delivery such as text documents, images, audio, or video. Information can be accessed wherever we are in real-time, anywhere, and any place. A search Engine is a medium that is used in finding various information on the internet. The phenomenon of problems that arise in the search for information on the internet is the search results of the desired topic are often not relevant to the keywords entered. Information Retrieval is a system, method, and procedure used to recover information stored from a collection of information based on a query entered by the user. The performance of searches based on Information Retrieval can be improved by various methods. One of them uses the Query Expansion method, which works by reformulating the initial query by adding several terms using Thesaurus to the query entered so that it is expected to be able to increase the relevance of data obtained from search results. The contribution obtained from this research is the development of an Information Retrieval System-based document search engine that can improve data relevance and the quality of document search results. Based on the test results using the recall and precision methods, a graph was obtained showing that the relevance results with the Query Expansion method had increased in the level of data relevance. Because when using expansion queries, the recall results are higher, allowing more relevant documents to be retrieved.
Keywords:	Information Retrieval, Query Expansion, Thesaurus, Data Revelation, Process Innovation
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	INNOVATIVE TIME SERIES-BASED ECG FEATURE EXTRACTION FOR HEART DISEASE RISK ASSESSMENT
Author:	A. DEEPAK KUMAR, N. REVATHI, S. IRIN SHERLY, R. LALITHA, R. VINSTON RAJA
Abstract:	Early detection of heart disease is crucial in reducing the mortality rate caused by cardiac events. Electrocardiogram (ECG) signals are widely used in clinical practice to diagnose various heart diseases. However, the physical analysis of ECG signals by experts is time-wasting and subjective, leading to a need for automated methods for ECG signal investigation. For the automatic classification of ECG data, numerous feature extraction algorithms have been put out recently. In this research, we provide a unique feature extraction approach for time series-based electrocardiogram (ECG) signals to predict the risk of heart disease. The proposed approach combines wavelet transform and principal component analysis (PCA) for the derivation of the discriminative features from ECG signals. The extracted features are then fed into a machine learning model for heart disease prediction. Using a publicly accessible dataset, the performance of the suggested strategy is assessed and contrasted with cutting-edge feature extraction techniques.
Keywords:	Electrocardiogram, Feature Extraction, Principal Component Analysis, Wavelet Transform.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	SAFETY RISK MANAGEMENT BOW-TIE ANALYSIS AND SAFETY PROMOTION IN THE OPERATIONS OF SMALL UNMANNED AIRCRAFT SYSTEMS
Author:	ARDHAN FATURACHMAN, DWI LESTARY, AGUSTONO, ARI SATRIA
Abstract:	SUAVs can be post a threat to aircraft because they can cause fatal accidents. Even though its size is much smaller than an airplane, the presence of UAVs in the airport area is strictly prohibited. This study aims to determine the management of risks rising from the operation of SUAVs at Flight Safety Area (KKOP) at Soekarno Hatta International Airport. The methodology used is quantitative. Therefore, the authors conclude that risk management is very necessary in the world of aviation, because it concerns mutual safety, and with the increasing number of SUAVs with various types and weights, the authors hope that the level of safety awareness will also be higher for SUAVs users, also called remote pilots.
Keywords:	Safety Risk Management, Safety Promotions, Civil Aviation Safety Area, Small Unmanned Aerial Vehicle, Bow-Tie Analysis
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	A SYSTEMATIC LITERATURE REVIEW ON DIGITAL TRANSFORMATION PROJECTS: DISCOVERING PROJECT GOALS AND PROBLEMS
Author:	ADI AZHARI, TEGUH RAHARJO
Abstract:	Digital transformation has become a prevalent term in today's business landscape, signifying the adoption of technology to improve existing processes. However, not all digital transformation projects yield successful outcomes. Hence, it is crucial to identify the desired outcomes before implementing such projects. Organizations must prepared to tackle the challenges associated with project implementation. This study utilizes a systematic literature review to gather data on the goals and problems encountered during digital transformation initiatives. This study classifies project goals into different domains, namely operational, customer, social, financial, and strategic domains. Similarly, the problems and solutions are grouped into eight project performance domains: stakeholders, planning, team, project work, development approach and life cycle, uncertainty, delivery, and measurement. This research aims to provide organizations with insights into the necessary preparations for successful project implementation. Moreover, the study bridges the gap between theoretical concepts presented in books and real-world case studies, enhancing understanding in this field.
Keywords:	Digital Transformation, Project Goals, Problems and Solution, Project Management, Technology Adoption
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	SHAPING TRUST AND LOYALTY IN ONLINE COMMERCE: AN EMPIRICAL STUDY OF INFLUENTIAL FACTORS
Author:	JUAN DANIEL WIJAYA, MAHESWARA RABBANI, ANGELINA GRACIA EDDYPUTRI BURHAN, ASTARI RETNOWARDHANI
Abstract:	There have been reported cases related to cybercrime carried out in E-commerce technology, with various types and methods of cybercrime used by these cybercriminals, such as identity theft and data transactions, unauthorized use of personal data, phishing and click-baiting, and payment fraud. This research aims to test a hypothesis centered on the impact of cybercrime and trust in using E-commerce technology from the consumer or user side. This research contributes to identifying what factors can affect user trust such as perceived qualities, privacy, security and satisfaction so that E-commerce implements its strategy to increase user loyalty. The survey was distributed to 444 respondents who had used e-commerce for shopping. The research results show that user satisfaction is influenced by information quality, system quality, and service quality. Trust is influenced by security, privacy, and satisfaction. The loyalty variable is influenced by trust. Meanwhile, satisfaction is not influenced by security and privacy.
Keywords:	E-commerce, Indonesia, Trust, Quality, Cybercrime, Loyalty
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	A RANDOM EARLY DETECTION (RED) TECHNIQUES TO OVERCOME THE CONGESTION PROBLEM IN VEHICULAR ADHOC NETWORKS (VANETS)
Author:	MOHAMMAD RAMEZ ALI, MOHD FADZIL ABDUL KADIR, AHMAD FAISAL AMRI ABIDIN, MOHAMAD AFENDEE MOHAMED, ABD RASID MAMAT, ZURADZMAN MOHAMAD RAZLAN, ZUNAIDI IBRAHIM
Abstract:	A vehicular ad-hoc network, also known as VANET, is described as one of the most challenging domains when providing an intelligent transport system (ITS). VANET is a critical application area for the mobile ad-hoc network also regarded as MANET. However, the effectiveness of VANET depends on the ability to detect congestion and resolve security issues to ensure effective and efficient services are provided to the users. This paper aims to examine random early detection techniques applicable in determining congestion issues in VANETS and propose the working framework useful for handling congestion in VANET via the provision of effective queue management approaches. The approach presented in this study paper is known as Node Based Throughput (NBTH), which focuses on evaluating the throughput of nodes. In this case, the approaches were examined and scrutinized with different parameters, such as end-to-end throughput delays and packet loss associated with the density nodes.
Keywords:	VANET, MANET, Congestion, vehicular network.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	FORECASTING ELECTRICITY CONSUMPTION THROUGH A FUSION OF HYBRID RANDOM FOREST REGRESSION AND LINEAR REGRESSION MODELS UTILIZING SMART METER DATA
Author:	DR. B. RAVI PRASAD, MR. DUDEKULA SIDDAIAH, PROF. TS. DR. YOUSEF A.BAKER EL-EBIARY, DR. S. NAVEEN KUMAR, DR. K SELVAKUMAR
Abstract:	This project seeks to forecast energy usage by utilizing data from smart meters and the Random Forest Regressor algorithm. A logical flow of processes includes data preparation, model training, and model performance assessment. The use of the Label Encoder method to the dataset initiates the conversion of category variables into numerical representations. This crucial phase guarantees that the model can skillfully handle the data processing. After that, missing data are handled using the Simple Imputer technique, which judiciously fills in the blanks with suitable measurements like mean or median values. The train_test_split function divides the dataset into training and testing subsets, preparing the way for model training. The Hybrid Random Forest Regressor approach is used in combination with the LR methodology to train the predictive model. Numeric characteristics are standardized using the Min Max scaling approach, which aligns them into a common range, to ensure the best model performance. A wide range of evaluation measures, such as mean_absolute_error, mean_squared_error, and median_absolute_error, are used to evaluate the model's effectiveness after training. These metrics provide a lot of insightful information by measuring the precision and accuracy of the model's forecasting abilities. The Random Forest Regressor algorithm, together with various preprocessing techniques, allows this research to forecast energy use from smart meter data with a high degree of accuracy. A spectacular Mean Absolute Error of proposed method is 70.79, outperforms over existing methods, SARIMA and SVR+LR and an equally excellent Median Absolute Error of 30.46 are achieved by the resulting model. The proposed model is implemented using Python software. These error rates provide quantifiable benchmarks that reveal the model's performance features and are an indication of its extraordinary predictive precision. The study's findings have enormous potential for improving cost effectiveness, eco-aware practices, and energy management effectiveness.
Keywords:	Smart Meter Data; Energy Consumption; Random Forest Regressor; Simple Imputer; Min Max Scaler
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	ELITE ARTIFICIAL BEE COLONY OPTIMIZATION-BASED SYNERGY RANDOM FOREST (EABC-SRF) FOR ADDRESSING AMBIGUITY IN STOCK TWEETS
Author:	G.PRIYADARSHINI, Dr.D,KARTHIKA
Abstract:	The stock market is a complex and dynamic financial ecosystem where investors buy and sell securities. Various factors influence it, including economic indicators, geopolitical events, and social sentiment. Twitter has become a significant source of real-time information for traders and investors. Stock tweets are short messages posted on Twitter that discuss stocks, providing insights, opinions, and predictions. Analyzing these tweets can help gauge market sentiment and anticipate price movements. The classification of stock tweets involves categorizing them as positive, negative, or neutral based on sentiment. This sentiment analysis aids in understanding investor sentiment and predicting market trends. The Elite Artificial Bee Colony Optimization-Based Synergy Random Forest (EABC-SRF) is an innovative algorithm to enhance sentiment analysis. It combines the power of Artificial Bee Colony Optimization (ABC) and Synergy Random Forest (SRF) to optimize feature selection and sentiment classification. EABC-SRF uses elite artificial bees to select the most relevant features for sentiment analysis. It then integrates these features into the SRF framework to classify tweets effectively, reducing ambiguity and noise. The “Stock Tweets for Sentiment Analysis and Prediction” dataset is the foundation for training and testing EABC-SRF. It contains a vast collection of stock-related tweets for model development and evaluation. Results from experiments with EABC-SRF demonstrate its superior performance in sentiment analysis compared to traditional methods. It disentangles the ambiguity in stock tweets, providing valuable insights for investors and traders in predicting market sentiment and trends.
Keywords:	Stock Market, Twitter, Classification, Random Forest, Artificial Bee Colony, Optimization
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	AUTOMATED MUTATION ANALYSIS FOR SMART CONTRACT USING AMA TOOL WITH ENHANCED GA AND MACHINE LEARNING APPROACH
Author:	R SUJEETHA, K AKILA
Abstract:	Smart Contracts are the critical most popular in the Dapps Blockchain network. Smart contracts play a vital role in safety-critical products. The quality of the smart contract is a vital factor. Test cases are used to ensure the correctness of the smart contract code. The efficiency of the smart contract test suite is assessed using the mutation testing technique. The state-of-the-art tools for assessing test suite quality generate numerous mutants for execution. The test case generation-related state-of-the-art tools code and functional coverage require further research to provide better coverage. This article proposes a tool for performing automated mutation analysis (AMAT) for smart contracts in which the test cases are generated using the proposed enhanced GA used for the mutation analysis. The mutation testing utilizes the effective mutants obtained using a machine learning-based classification algorithm for reducing the number of mutants executed. The results show that the tool effectively generates optimized test cases with high branch and function coverage and achieves up to 98% mutation scores.
Keywords:	Test Case, Genetic Algorithm, Mutation Testing, Classification
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	PERFORMANCE COMPARISON OF K-NEAREST NEIGHBOR AND NAIVE BAYES ALGORITHMS FOR HEART DISEASE PREDICTION
Author:	ROHMAT TAUFIQ, ANGGA ADITYA PERMANA, ARSANAH1, ANALEKTA TIARA PERDANA
Abstract:	Heart disease is leading causes of death in several countries, including Indonesia. Accurate and precise prediction of heart disease is so important. There are lots of machine learning algorithms that can be used to make predictions. In this study, 2 classification algorithms: K-Nearest Neighbors and Naive Bayes were used to compare the performance of algorithms for better prediction. Heart disease dataset UCI Machine Learning Repository Center for Machine Learning and Intelligent Systems dataset was used for generating confusion matrix values, analyzing accuracy values in predicting heart disease based on 14 attributes. The results showed that the KNN algorithm has an accuracy value of 91.25% while Naive Bayes is 88.7%.
Keywords:	KNN, Naïve Bayes, Comparison, Heart Disease
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	FULLY CONVENTIONAL CASCADED MULTI DIMENSIONAL NETWORK FOR COLORECTAL POLYP SEGMENTATION OF COLONOSCOPY IMAGES
Author:	M.N. PRASHANTH, S. SENTHILKUMAR
Abstract:	Early polyps detection can reduce the risk of colorectal cancer. Segmentation of colonoscopy images can facilitate a faster diagnosis of polyps. However, accurate polyp image segmentation is a difficult process because the size, shape, and location of polyps vary, and expert experience directly affects this process. This research article presents a novel method for the automatic segmentation of polyp areas from colonoscopy images. Specifically, to accurately identify the polyp boundaries, we use a hybrid LGB color space in which the primary colors green and blue with the Lightness component of CIE-L×a×b are concatenated. . In this research work, Multi-Dimensional Cascades Network (MDCNet) is developed for the early stages of Colorectal Polyp Segmentation. It consists of two stages, stage 1 involves performing polyp location and rough segmentation using a probability-anatomical-prior-guided shallow-layer enhanced 3D location net. In order to lower the uncertain probabilities and false positive boundary points, a novel circular inference module and parameter Dice loss are also suggested. A multi-view 2.5D net made up of three 2D refinement subnetworks is used in stage 2 to thoroughly investigate the morphological features, making up for errors and missing spatial information of a single view. The proposed MDCNET-based segmentation model detects the early stage of polyps of colonoscopy images. The experimental results are validated using the Sensitivity, PPV, Dice similarity coefficient(DSC), and IOU values. The DSC and IOU values for the segmentation results of the proposed method are 0.9259, and 0.8938 in the Kvasir-SEG Data set and 0.8109, 0.8045 in the CVC-ClinicDB dataset respectively.
Keywords:	Colorectal Cancer; Polyp Segmentation; Fully Convolutional Network; Cascading Mechanism.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	ASSESSING SDG 11.3.1 THROUGH MACHINE LEARNING-BASED CLASSIFICATION OF LANDSAT DATA AND DASYMETRIC POPULATION MAPPING: KENITRA CITY CASE STUDY
Author:	SAADANI MOUSSA , ELBRIRCHI ELHASSAN, BACHIR ALAMI OMAR
Abstract:	In the current era, where the primary focus is on achieving sustainable urban development, the uncontrolled expansion of urban areas poses a substantial threat to their long-term sustainability. Within this context, Goal 11 of the Sustainable Development Goals (SDGs) assumes critical importance. This specific objective is dedicated to evaluating the sustainability of urban progress, with Indicator 11.3.1 serving as a central metric, measuring the "Ratio between the rate of land consumption and population growth." Faced with the need to quantify this indicator and recognizing the limitations inherent in traditional data sources, a shift toward non-conventional data becomes imperative. This transition to advanced methodologies is increasingly vital for a more accurate and comprehensive understanding of urban dynamics. The study focuses on a comprehensive assessment of SDG Indicator 11.3.1 in Kenitra city, spanning multiple time points (1994, 2004, 2014, and 2022). By utilizing Landsat imagery in conjunction with machine learning classifiers and Dasymetric population mapping, the research yields illuminating insights into the ratio of Land Consumption Rate to Population Growth—a crucial measure for assessing sustainable urbanization. From 1994 to 2004, this ratio stood at 2.64, indicating significant land consumption compared to population growth. Over the subsequent decade (2004-2014), the ratio decreased to 2.25, signifying a more balanced expansion trajectory. Particularly noteworthy is the period between 2014 and 2022, marked by a substantial decrease in the ratio to 0.68, reflecting a significant shift toward optimal land use. These findings highlight the dynamic evolution of urban development dynamics and the importance of strategic approaches for promoting sustainable urban growth.
Keywords:	Sustainable Development Goals, Landsat, Google Earth Engine, Machine Learning Classifiers, Dasymetric Population Mapping, Morocco, Kenitra
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	REPRODUCTION OF COMBINED EFFECTS ON ECOLOGICAL SYSTEMS AND THEIR COMPONENTS IN SIMULATION MODELS
Author:	SERGEY MAMIKHIN, ALIYA BUGUBAEVA, DENIS LIPATOV, DMITRIY MANAKHOV, TATYANA PARAMONOVA, VALERIYA STOLBOVA, ALEKSEY SHCHEGLOV, VADIM CHASHKOV
Abstract:	The study aimed to develop an algorithm for reproducing the simultaneous impact of various factors, both biogenic and anthropogenic, on ecological systems and their components in simulation models. Simulation models are effective tools in predicting and understanding the complex interactions between biogenic and anthropogenic factors in ecological systems that help to develop conservation strategies and sustainable resource management. The authors constructed the EcoCombi model, which reproduced the dynamics of the number or biomass of living components of biological systems that are under the combined influence of two or more factors. The object of the simulation was a terrestrial ecosystem consisting of the following components: soil, herbaceous vegetation, herbivores, and predatory animals. As a result, an algorithm was developed for reproducing the multifactorial impact on the object, based on which simulation models can be built and implemented, allowing for the prediction or reconstruction of various environmental situations associated with the simultaneous impact of various factors, both biogenic and anthropogenic, on ecological systems and their components. Based on the results of numerical experiments, it was found that an exhaustive account of the object's properties played a key role in an effective assessment of the consequences and forecasting of the combined effects of factors on the object using a model. It is also important to consider the nature of the impact of the factor (mono- or poly-impact on the vital processes of organisms, single or chronic exposure). The proposed algorithm applies to any situation where there is a multifactorial impact on the object. In this respect, it can be effective for reproducing most of the environmental processes observed during technogenic pollution.
Keywords:	Factorial Ecology, Combined Impact, Simulation Modeling, Ionizing Radiation, Ecotoxicants.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	DEVELOPING A MIXED NONPARAMETRIC REGRESSION MODLLING (SIMULATION STUDY)
Author:	AMJED MOHAMMED SADEK, LEKAA MOHAMMED ALI
Abstract:	The relationship between the response variable and the explanatory variable can show discernible patterns in certain cases, while in others, such patterns can remain unknown. Nonparametric regression techniques can potentially be employed to ascertain the unidentified pattern of relationships. The nonparametric regression technique provides a high degree of flexibility. In this work, we propose a new function for the kernel and use it with the mixed nonparametric regression between the kernel and truncated spline to compare it with the mixed Gaussian and the mixed biweight. Therefore, according to the study performed, we suppose that have a specific pattern that was used with the spline method. On the other hand, do not have a specific pattern that was used with the kernel method. Based on the GCV and MSE values, the best model was produced using the optimal bandwidth for each variable and one point of optimal knot for various sample sizes. The simulation study demonstrated that the mixed model with the proposed function has a suitable and superior performance when compared to the mixed Gaussian and mixed biweight models. The results of this investigation clearly demonstrated this model's superiority over its competitors. The highest obtained results have been confirmed by the coefficients of determination (R-square), which are 90.5%, 94.4%, and 97.2%, and the mixed nonparametric model with suggested function (AMS) provided the lowest mean square error (MSE) values of 4,074, 2,185, and 2,361 for various sample sizes. This indicates that the model will be able to produce accurate predictions and improve the performance of the data that we have been concentrating on.
Keywords:	Kernel Regression, Spline Truncated, Nonparametric Regression, Mixed Estimator, Mean Square Error.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	GENERATION OF LOSSLESS DWTS FOR SPECTRAL FEATURE EXTRACTION USING DISCRETE WAVELET 2D-CNN MODEL FOR CLASSIFICATION OF HYPER SPECTRAL IMAGES
Author:	SARITHA HEPSIBHA PILLI, VALLI KUMARI VATSAVAYI
Abstract:	In hyper spectral imaging (HSI), sensors capture detailed spectral data across numerous narrow spectral bands, resulting in high dimensionality. This dimensionality issue significantly affects HSI classification. Therefore, feature extraction (FE) for dimensionality reduction is crucial in HSI processing. This study explores the use of a 2D-Convolutional Neural Network (CNN) for HSI data analysis. Traditional 2D-CNNs, however, may not effectively integrate spectral and spatial features for HSI classification. To enhance the CNN’s architecture for HSI categorization and analysis, this research investigates the integration of a lifting-scheme-based Discrete Wavelet Transform (DWT) with 2D-CNN, which we refer to as the Discrete Wavelet 2D-CNN model. The proposed methodology’s main objective is to provide guidance for future research in selecting the appropriate mother wavelets for spectral FE in conjunction with a 2D-CNN classifier. By integrating 2D-CNN with the DWT, which maintains spectral signature distinctions, it may enhance the emphasis on spectral features. The novelty of this work lies in its accuracy evaluation of three DWTs: Haar, Daubechies 4-tap orthogonal filter (D4), and Cohen-Daubechies-Feauveau 9/7-tap biorthogonal filter (CDF-9/7-wavelet), for spectral FE in HSI classification using Discrete Wavelet 2D-CNN. This approach utilizes a lifting scheme-based DWT for spectral FE in HSI. The lifting-scheme is an effective nonlinear transformation approach for DWTs. The resulting spectral characteristics from wavelet decomposition are then fused with a 2D-CNN, preserving spatial information, thus creating a spectral-spatial feature vector for classification. The Discrete Wavelet 2D-CNN model's accuracy was assessed on benchmark HSI datasets, including Indian Pines (IP), Salinas (SA), and Pavia University (PU). It was observed that the D4 wavelet-based model outperformed other configurations. Furthermore, we compared the model's classification accuracy with several state-of-the-art deep learning algorithms and found that the choice of the mother wavelet for HSI spectral FE can significantly impact the model's overall performance.
Keywords:	Hyper Spectral Images (HSIs) Classification, Discrete Wavelet Transforms (DWTs), Convolutional Neural Network (CNN), Lifting-Scheme, Feature Extraction (FE)
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	ADVANCING REAL-TIME VIDEO VIOLENCE DETECTION: A DEEP LEARNING APPROACH WITH INTEGRATED TELEGRAM ALERTING
Author:	IMANE RAHIL, WALID BOUARIFI, OUJAOURA MUSTAPHA, RAHIL GHIZLANE
Abstract:	In todays security and monitoring landscape, remote surveillance has become a cornerstone. However, its conventional role as a passive historical event recorder has restricted its potential as a proactive detection tool. In response to this limitation, we present an innovative paradigm shift powered by deep machine-learning techniques. At the heart of our pioneering approach lies the utilization of Convolutional Neural Network (CNN) models, celebrated for their prowess in image analysis. These models are instrumental in extracting pertinent features from the observed images, forming the bedrock for a more exhaustive analysis. These features are methodically organized into temporal sequences, constructing a chronological depiction of event progression. To elevate our approach, we seamlessly integrate recurrent models known for their proficiency in analyzing sequential data and unveiling patterns over time. This fusion enables refined, real-time event analysis, a departure from the conventional practice of recording for later analysis. The synergy of deep machine learning and temporal sequence processing offers a promising strategy to amplify the effectiveness of remote surveillance, transcending the boundaries of traditional methods by providing a holistic understanding of the observed events. This empowerment allows the system to identify meaningful patterns and behaviors, significantly enhancing real-time detection capabilities. Furthermore, our approach underscores the significance of immediate alerting, bridging the gap between detection and response. Through the incorporation of a Telegram bot, the system can proactively transmit real-time alerts, ensuring swift action in response to emerging events. Additionally, our system's capacity to capture and extract facial information from video data enhances our understanding of individuals involved in violent incidents. This supplementary layer of information proves invaluable for subsequent investigations and the tracking of potential threats. This innovative approach symbolizes a revolutionary shift in remote surveillance. By amalgamating deep machine learning, temporal sequence processing, and real-time alerting, our goal is to overcome current limitations, empowering surveillance systems to actively recognize and respond to ongoing incidents in real-time. This transformation harbors the potential to fortify security measures, expedite emergency response times, and amplify overall public safety. Through our approach, we envisage a future where remote surveillance systems metamorphose into intelligent, proactive guardians, safeguarding the well-being of individuals and communities.
Keywords:	Violence detection, Deep learning, Machine learning, Real-time detection, Face detection
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	DEEP LEARNING IN STOCK MARKET PREDICTION: A FIVE-YEAR LITERATURE REVIEW ON DEVELOPMENTS, CHALLENGES, AND FUTURE DIRECTIONS
Author:	BERNADECTUS YUDI DWIANDIYANTA, RUDY HARTANTO, RIDI FERDIANA
Abstract:	Shares or equities have received significant attention in investment because of their profit potential. However, with the complexity and volatility in the stock market, the need arises for more accurate prediction methods. In the last decade, Deep Learning algorithms have become a promising solution. Deep learning methods offer superior capabilities in handling big data and non-linear relationships. This research reviews the development of Deep Learning algorithms in stock market predictions from 2017 to 2022. This research uses a literature review methodology. This research reviewed 86 articles selected from the initial 346 articles for further analysis. The analysis results show the dominance of using historical trading data as system input in stock price predictions. The New York and Shanghai Stock Exchanges are the main focus of researchers' attention. This research also identifies the potential for combining input data and optimizing prediction methods as a future research opportunity. This research generates opportunities to develop algorithms based on LSTM and Bi-LSTM, hybrid and ensemble methods. It is hoped that this research can provide insight into the latest developments in stock prediction using Deep Learning and provide insight into future research directions.
Keywords:	Stock Market Prediction, Deep Learning, Literature Review, Feature Engineering, Stock Forecasting
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	A NOVEL DEEP LEARNING MODEL FOR RAISIN GRAINS CLASSIFICATION
Author:	TAHA M. MOHAMED, SARAH N. ABDULKADER
Abstract:	Automatic goods classification is important to facilitate the operations performed at customs. Raisin grains are one of these important goods that need accurate classification. The performances of the algorithms available in the literature are not satisfactory. So, this paper presents a high performance deep-learning model for raisin grains classification. The proposed model performs the necessary preprocessing steps on a publicly available dataset. The architecture of the proposed model is the standard architecture which is more appropriate for the tabular dataset used. Many preprocessing tasks are evaluated before constructing the deep network. Moreover, the features of the dataset are analyzed and preprocessed by using feature normalization and principal component analysis (PCA). The performance of the deep network is evaluated using different network configuration for optimal dataset modeling. Moreover, the proposed model evaluates the validity of some network regularization techniques to maximize the performance of the proposed model. The proposed results show that the proposed model outperforms other works in the literature when applied on the same dataset. The proposed accuracy and F-Score exceed 91% in high correlation dataset features. Comparing to previous works, at least 5% improvements are achieved using the proposed model. An important conclusion from the obtained results is that dimension reduction methods are not effective on all datasets. The results also support the importance of the preprocessing steps in enhancing the deep network performance.
Keywords:	Deep Learning; Raisin Grains Classification; Hyper-Parameters Tuning; Features Normalization.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	EFFICIENT VEHICLE ROUTING IN CONGESTED REGIONS
Author:	IBRAHEEM M. ALHARBI
Abstract:	Optimization methods could optimize costs and determine the optimal resources needed for supply chain management (SCM). The vehicle route selection problem in congested regions is a main challenge in SCM. The problem is to determine the daily shortest optimal route among several routes in congested regions. Unfortunately, optimization techniques could only find static fixed routes. However, finding dynamic routes, suitable for congested regions, is still challenging. In this paper we will propose a model for solving vehicle routing problem in congested regions. The proposed model combines both optimization techniques with live traffic information to determine daily optimal routes. Also, the paper will discuss the challenges of using both static and dynamic methods in solving vehicle routing problems in congested regions. The addressed problem has many applications in different fields especially in supply chain management, crowd management, fleet management, and congested cities.
Keywords:	Global Optimization Methods, Live Google Maps, Supply Chain Management, Fleet Management, Travelling Salesperson Problem, Artificial Intelligence, Vehicle Routing Problem.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	HEURISTIC LEARNING APPROACH FOR RELIABLE MULTI HEAD COMMUNICATION IN WIRELESS SENSOR NETWORK
Author:	MURALI YACHAMANENI, RADHIKA BASKAR
Abstract:	In recent time, Wireless sensor network has gained a large usage in many Real-time applications. The capability of exchanging data with no pre infrastructure support has given a boost for the usage of wireless sensor network (WSN) in various practical application. Sensor nodes interfaced in WSN are inter-linked to each other and exchange data via different mode of communication wirelessly. Most popular mean of data exchange in WSN is a cluster-based communication, where large monitoring area is sub divided into small clusters for exchanging data. Cluster based communication involve head nodes, which act as a major link point in exchange of data. Optimal Selection of head node in WSN is a critical task for efficient communication. Heuristic learning approach such as Reinforcement learning is proposed for optimal selection of head node in WSN. Existing learning methods were developed with the objective of offering high throughput with energy conservation in the network. However, the reliability of the head node in data exchange is not addressed. Reliability factor is observed to be an important factor in WSN because of the sensitiveness of senor data in the network. For developing a reliable head node selection in WSN, this paper contributes in developing a new ‘Reliable Head Node Routing’ (RHNR) approach for optimal head selection in WSN with reliability factor consideration. Proposed RHNR method process on observatory metric of data forwarding at each node in the network and develop a reliability factor updating existing reward factor for Head selection in WSN. The existing Reinforcement approach using Q-Learning is updated with the metric of forwarding factor in selection of a reliable head node in WSN communication. Simulation observations of developed method obtained increase in network throughput and node life time with varying node counts and payload in the network. The End-to-End (E2E) delay metric decreased with the proposed RHNR method resulting into a faster and more reliable packet exchanging in WSN communication.
Keywords:	Wireless Sensor Network, Heuristic Learning method, Reliable Head Node routing, Multi Head Communication.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text

Title:	ANALYSIS OF VARIATION OF FEATURE EXTRACTION METHODS IN THE CLASSIFICATION OF AL-QUR AN MAQAM USING MACHINE LEARNING
Author:	MUHAMMAD AHMAD AGIL ALAYDRUS, AMALIA ZAHRA
Abstract:	As the book of religion with adherents of more than 1.9 billion people in the world and 86.7% of Indonesia's population, the Qur'an has earned the title of being the 'book' that is most widely read in the world. Al-Qur'an is the holy book that guides Muslims in life and religion. The Koran is read every day by a Muslim, for example during prayer services. Until now, the process of learning and memorizing the Qur'an has never stopped; the development of applied technology to support the learning process is also intensively carried out. Until now, studies that have focused on reciting the Qur'an have not made maqām the main focus. In contrast to recitation, discussions and learning applications can be found on various platforms in various forms. Tajweed does indeed play an important role as a basis for reciting the Koran, but limited information and support systems in maqām learning often prevent someone from learning it. This is the background of this research to move in presenting a system that can assist the maqām learning process. This classification system was built with the hope that maqāmat learners can detect the types of maqām independently, both from personal readings and voice recordings of other people. Thus, users can verify whether the chanting tone of the Al-Qur'an recitation is appropriate, or find out the maqām from someone elses reading recording.
Keywords:	MFCC, SDCC, Maqam, and CNN
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2023 -- Vol. 101. No. 21-- 2023
Full Text