		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
September 2024 | Vol. 102 No.18

Title:	HAND GESTURE CLASSIFICATION USING TIME-FREQUENCY IMAGES AND VALIDATION APPROACHES
Author:	CHAPARRO MACIAS CAMILO ANDRES, CARDOZO SANCHEZ MAICOL ESTIVEN, JUAN PABLO GASCA CALDERÓN, DIEGO FERNANDO GONZÁLEZ, RUTHBER RODRÍGUEZ SERREZUELA
Abstract:	Surface electromyographic (sEMG) signals are a non-invasive method for acquiring signals that play a fundamental role in the monitoring of prosthetic devices by providing information about human motor functions. This leads to the need for accurate classification of sEMG signals, despite variations in signal stationarity, the presence of sensor noise, differences between the muscles involved, and the peculiarities of each patient. This study focuses on the classification of hand grip postures using sEMG signals acquired from amputee patients. Special emphasis is placed on the use of the time-frequency domain for feature extraction, using the spectral analysis of the reduced-time Fourier transform (STFT). To carry out this task, a classification model based on a convolutional neural network (CNN) is used. The classification method is adjusted, trained, and evaluated through three experiments. The first, called "One to One", yields accuracy percentages of 90.84%, 91.05%, and 91.13% for spectrograms of 32x32, 64x64, and 128x128 in size, respectively. In the second validation, called "All by One", an accuracy of 62.28% is achieved for spectrograms of 32x32 pixels. Finally, in the last K-fold cross-validation validation, an average accuracy of 86.73%, 86.77%, and 87.97% is obtained for spectrograms of 32x32, 64x64, and 128x128 in size, respectively.
Keywords:	Electromyography, Hand Gesture, Classification, STFT, CNN.
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	ANALYSIS OF MACHINE LEARNING METHODS FOR GENDER AND AGE IDENTIFICATION
Author:	NUR NAFIIYAH, AYU ISMI HANIFAH, EDY SUSANTO, EHA RENWI ASTUTI, RAMADHAN HARDANI PUTRA, ENDANG SETYATI
Abstract:	An automatic individual identification system is needed to support the forensic odontology process more efficiently and easily because there is still opportunity to be developed. The purpose of this research was to analyze the machine learning method for gender and age identification based on mandibular parameters in panoramic radiography. The machine learning methods used are MLP (Multilayer Perceptron), Decision Tree, Naive Bayes, k-NN (Nearest Neighbors), Logistic Linear, and SVM (Support Vector Machine). The data for this research were taken from the Dental and Oral Hospital, Faculty of Dentistry, Universitas Airlangga Surabaya. The data consisted of 120 patients based on the validation results of radiology experts, consisting of 61 males and 59 females, and was divided into 104 training data, and 16 testing data. The mandibular image on panoramic radiography was measured for nine parameters, namely ramus height left , ramus height right , ramus length left , ramus length right , bigonial width , bicondylar breadth , anterior mandibular corpus height left , anterior mandibular corpus height right , mandibular corpus length using the ImageJ application by radiology experts. The best machine learning method for gender identification is k-NN, with evaluation values of accuracy, precision, recall, and f1 score, respectively, of 0.750, 0.764, 0.750, and 0.733. And the best method for age identification is MLP, with values of accuracy, precision, recall, and f1 score, respectively, of 0.625, 0.267, 0.350, and 0.297.
Keywords:	Mandibular Parameters; Machine Learning; Gender Identification; Age Identification.
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	CHI SQUARE FEATURE SELECTION FOR IMPROVING SENTIMENT ANALYSIS OF NEWS DATA PRIVACY TREATS
Author:	DEFITROH CHEN SAMI UN, ARIS SUGIHARTO, FERRY JIE
Abstract:	Data security and privacy issues are becoming increasingly pressing in the technology-driven digital era. In 2022, this issue became a major topic in Indonesia and triggered various responses on social media. YouTube, one of the primary platforms, plays a crucial role as a news source. To understand public reactions to this news, sentiment analysis is employed as a research method. The initial stage before conducting sentiment analysis involves data preprocessing, which includes cleaning, case folding, tokenization, slang correction, stemming, and stopword removal. Subsequently, the TF-IDF method is used to assess the significance of words in documents, and Chi-Square feature selection is applied to enhance the performance of the classification model. The main contribution of this study lies in the application of Chi-Square feature selection to improve sentiment analysis accuracy in the context of data privacy threat news. Chi-Square feature selection has proven to be effective in identifying the most relevant features, thereby eliminating irrelevant features and enhancing the accuracy of the classification model. The use of the C5.0 algorithm combined with Chi-Square feature selection achieved the highest accuracy of 87.34%, compared to the 80.14% accuracy achieved without the Chi-Square feature selection method. This research makes a significant contribution by demonstrating that appropriate feature selection methods can substantially improve sentiment analysis model performance, providing a more accurate and effective approach to managing and analyzing sentiment data from social media platforms.
Keywords:	Privacy data, YouTube, Sentiment analysis, Chi-square feature selection, C5.0 algorithm
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	OPTIMIZING ENERGY EFFICIENCY IN WIRELESS AD HOC NETWORKS FOR DISASTER MANAGEMENT: A NOVEL BROADCAST APPROACH FOR PROLONGED SENSOR LIFE
Author:	FAISAL ALZYOUD, MONTHER TARAWNEH, IBRAHIM ALTARAWNI, MOHAMMED AMIN ALMAIAH, RAMI SHEHAB, TAYSEER ALKHDOUR, ROMEL AL-ALI6 AND THEYAZAN H.H ALDAHYANI
Abstract:	Globalization, industry, population number growth are considered the main reasons for climate changes, so when some nature calamity or disaster occurs in any area, it leads to isolate this area from the rest of the region. Wireless ad hoc networks are used in disaster management to reduce losses and cost as they reduce the response aiding time to save lives. However, they suffer from Limited energy which controls the lifetime of the sensor. Different methods applied to save energy and increase battery life such as sleep-awake schedule which assume similar initial energy for all nodes. Domestic partition in unit disk graph is the producing of maximum possible number of such disjoint dominant. However, it assumes different initial energy level for all nodes, which is another problem. In this paper, a new broadcast approach based on minimum set data broadcasting proposed to minimize the number of broadcasting messages to increase the battery life during disasters and enhance the rescue operations.
Keywords:	Wireless Ad Hoc Networks; Disaster Management; Broadcasting Messages; Prolonged Sensors.
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	POWERING UP EFFICIENCY: A DEEP LEARNING MODEL FOR ACCURATE ELECTRICITY CONSUMPTION FORECASTING
Author:	MOHAMED MAHMOUD HASAN, NEMAT EL-TAZI, RAMADAN MOAWAD, AMANY H. B. EISSA
Abstract:	The evolution of intelligent power methodologies has emerged recently. This evolution is based on extracting smart grids` information for electricity management. One of the key challenges in electricity management is forecasting consumption. Forecasting electricity consumption provides the ability to utilize resources and reduce costs efficiently. This paper proposes a novel hybrid deep-learning model for short-term electricity consumption forecasting that combines traditional consumption data with other external features. The proposed model utilizes a time series dataset, climatic features (temperature, wind speed, and humidity), and specific holiday information. These additional features are intended to improve the accuracy of electricity consumption forecasting, thereby enabling more efficient resource utilization and cost reduction. The data pre-processing phase includes adjusting time units and adding new features. The proposed model for processing the data begins with a multi-convolutional neural network (CNN) for feature extraction purposes. Then, these extracted features are passed through stacked gated recurrent units (GRU) for electricity consumption forecasting. An additional dropout layer is introduced to avoid overfitting. Experiments are carried out to apply the proposed model to the real dataset. The performance of the proposed model is measured using accuracy metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) to assess the deviation between actual and forecasted values. The experiments show that our proposed model outperforms the published results of other research using techniques such as Linear Regression (LR), Long Short-Term Memory (LSTM), CNN-LSTM, Bidirectional LSTM (Bi-LSTM), Stacked LSTM, and Stacked Bidirectional LSTM on the same dataset.
Keywords:	Electricity forecasting, Intelligent energy, Deep learning, Convolution neural network (CNN), Gated Recurrent Neural Networks (GRU).
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	EXTRACTING A MULTIDIMENSIONAL CONCEPTUAL MODEL FROM A DATA LAKE USING AN MDA APPROACH
Author:	LAMYA OUKHOUYA, ANASS EL HADDADI, BRAHIM ER-RAHA1, ASMA SBAI
Abstract:	Data warehouse design is based on a thorough analysis of an organization's operational data sources. These sources are then reorganized into conceptual models to enable multidimensional analyses. However, extracting a model with a unified multidimensional structure across all of these data sources presents some difficulties because it is necessary to have full documentation of data sources to perform this task. To overcome these issues, this article presents an approach to modernizing data warehouses using a data lake as a source of consolidating data from the organization's operational sources. Our approach begins by extracting the relational physical model from each data source, which is then integrated into the data lake using a domain ontology. This ontology helps detect duplicate elements in physical models and merge them into a unified relational model for the data lake. Finally, from this unified model, we extract the multidimensional conceptual model. This approach is automated by aligning with the model-driven architecture. We also validated our contribution with a prototype whose objective is to design a tool for the automatic extraction of the conceptual model from a data lake consolidating the data sources. Furthermore, a comparison between our prototype and a manual process carried out by a computer scientist revealed that our prototype simplifies the extraction task and saves significant time compared to the manual process.
Keywords:	Data Warehouse , Data Lake , Multidimensional Model, Metadata , MDA Approach.
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	COMPARATIVE STUDY OF LOCAL WISDOM COMPREHENSION IN SHORT STORIES BETWEEN COLLEGE STUDENTS AND AI CHATBOTS (CHATGPT AND GEMINI)
Author:	JUANDA, AZIS, SULASTRININGSIH DJUMINGIN
Abstract:	Although AI has great potential, its application in culture and education still faces various obstacles. Teachers and universities should utilize ChatGPT to support student learning and promote problem-solving skills. This study aims to compare the understanding of local wisdom between university students and AI chatbots (ChatGPT and Gemini) in short stories and evaluate the effectiveness of AI as a learning tool in that context. This study involved 240 university students from diverse backgrounds and two AI systems, ChatGPT and Google Gemini, to compare the comprehension ability of local wisdom. Using a local wisdom value test and quantitative descriptive analysis, the study compared students' and AI's understanding and applied an independent t-test to examine differences by gender, university, and domicile. The findings demonstrated that AIs possessed a deeper and more thorough comprehension of local wisdom compared to university students, reflected in significantly higher mean scores and statistically significant differences. No notable differences were found in understanding local wisdom based on students' gender or domicile. However, a significant difference was observed between students from the two different universities. AI is a potent learning tool for local wisdom studies, highlighting the educational environment's role. In literature education, integrating AI offers a more efficient and accurate method for teaching complex cultural and literary values, enhancing students' text analysis and interpretation of symbolism and cultural context.
Keywords:	Artificial Intelligence, Digital Humanism, Education Technology, Local Wisdom, Students
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	AMALGAMATE APPROACHES CAN AID IN THE EARLY DETECTION OF CORONARY HEART DISEASE
Author:	VENKATESWARA RAO CHEEKATI,Dr.KRISHNA MOHAN KAJA,RAJA RAO PBV, GARIMIDI SUBBARAO,Dr.AYYAPPA CHAKRAVARTHI M
Abstract:	Due to the fact that heart disease is the leading cause of death worldwide, it is critical to recognize it early. Artificial intelligence (AI) is a relatively new technology that is being heavily applied in a variety of fields, including biomedical care and disease prediction. Deep learning and machine learning are two examples of relatively new technologies that are being heavily applied in the fields of biomedicine, healthcare, and the early detection of disease. The goal of this study is to see if human coronary heart disease risk factors can be predicted using risk variables (CHD). In order to evaluate the effectiveness of prediction techniques like K-Nearest Neighbors, Binary Logistic Classification, and Naive Bayes, it is required to measure the accuracy and recall of each prediction method (BLC). Bundling and boosting are examples of ensemble modelling techniques are comparable to these methods of predicting the future. For the purpose of determining whether or not ensemble techniques can improve the accuracy of coronary heart disease prediction, a comparative analytical method was adopted. These patient data records for coronary heart disease total approximately 70,000 records and serve as a testing ground for the modeling methodologies that are currently being researched and developed. There is a 1.96 percent increase in accuracy between bagged models and their conventional equivalents. The improved models outperformed all other models by a wide margin, with an average AUC of 0.73. A combined accuracy of 75.1 percent was achieved by using the SVM, KNN, and random forest classifiers, which were regarded to be the most accurate. Utilizing data analysis and K-Fold cross-validation, the performance of the tested models was assessed
Keywords:	Heart Disease, Hybrid Modeling, Artificial Intelligence, Ensemble Method, Machine Learning And Coronary Heart Disease
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	TASK SCHEDULING FOR VIRTUAL MACHINE MIGRATION IN CLOUD COMPUTING USING ADAPTIVE REINFORCEMENT LEARNING
Author:	MOHAMMED ZIAUR RAHMAN, ANANDARAJ SHANTHI PICHANDI
Abstract:	The number of cloud users and their respective workloads is continuously increasing due to the inherent benefits of Cloud Computing (CC). Due to the rapid increase in the use of cloud services, the energy consumption of cloud data centres is dramatically increasing. An overloaded or underloaded Virtual Machine (VM) leads to enhanced response time and energy consumption due to suboptimal resource utilisation. Adaptive Reinforcement Learning (ARL) approach is proposed in this research based on Task Scheduling (TS) for VM migration in a CC environment to overcome this limitation. The proposed ARL approach signifies the advancement in cloud resource management by incorporating Ant Colony Optimization (ACO) for dynamic VM migration and TS optimization which offers efficiency and adaptability in cloud environments. The ACO is selected for task scheduling because of its distributed optimisation adaptability, capability to enhance resource utilization, and the VM migration performances. The ARL approach performance is evaluated with metrics of throughput, migration time, response time, load, energy consumption and resource availability. The ARL achieves a throughput of 2.33, migration time of 12.64ms, response time of 128.57ms, and energy consumption of 0.45W which is superior when compared to the Selection strategy based on correlation and utilization (SS-CAU).
Keywords:	Adaptive Reinforcement Learning, Ant Colony Optimization, Cloud Computing, Task Scheduling, Virtual Machine Migration
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	YOLO ALTERNATES GROUNDED GARBAGE TAXONOMY AND REPROCESSING ASSISTANT
Author:	MADHAVI DEVI LANKA, K. PRASUNA, G VAMSI KRISHNA, MURALI KRISHNA ATMAKURI, P. RAVI KUMAR, PRAVEEN TUMULURU
Abstract:	The increasing number of people on the planet has led to an increase in resource consumption and trash production, underscoring the pressing necessity of efficient waste management in order to protect the environment. Regretfully, the recycling sector continues to face difficulties, chiefly related to precise waste classification, which is essential to the recycling process. Manual sorting hinders the recycling process and adds to inefficiencies since it is frequently prone to mistakes owing to subjective human judgment. In addition, the workers' health is seriously jeopardized by the inherent risks of direct touch when processing hazardous items. To address these issues, we suggest a ground-breaking fix: The Recycling and Garbage Classification Assistant using YOLO V5-V7 Versions. This method aims to improve trash sorting precision by utilizing image classification techniques. YOLO version V7 stands out as the leader with notable improvements in accuracy. This creative method reduces health concerns associated with handling dangerous chemicals by hand while also streamlining garbage sorting procedures by utilizing cutting edge technology. The incorporation of YOLO versions V5–V7 is a crucial step in ushering in a new era of recycling processes that are accurate and efficient, which will greatly contribute to the overall objective of environmental sustainability.
Keywords:	YOLO, Garbage, Taxonomy,V5-V7, Reprocessing
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	PROACTIVE CYBER DEFENSE AND FORENSIC INVESTIGATION TECHNIQUES FOR DRONE OPERATION: A HOLISTIC APPROACH
Author:	ALBIA MAQBOOL, JIHANE BEN SLIMANE, NOUHA KHEDIRI, MOHAMED BEN AMMAR, AMANI KACHOUKH, AHMAD ALSHAMMARI
Abstract:	The integration of drones into various sectors, such as logistics, surveillance, agriculture, and emergency response, has revolutionized operational capabilities. However, this advancement has also exposed drones to significant cybersecurity threats, necessitating robust forensic investigation techniques. This paper presents a comprehensive approach to enhancing cyber defense mechanisms and forensic investigation methodologies specifically tailored for drone operations. Leveraging deep learning and machine learning techniques, our proposed framework aims to detect, mitigate, and investigate cyber threats targeting drones in real-time. The framework was developed and validated using a combination of publicly available datasets, including the DARPA UAV attack scenarios dataset and UNSW-NB15 network intrusion data, as well as data from controlled drone operation simulations that replicated real-world scenarios, such as surveillance and delivery missions under cyber-attack conditions. This comprehensive approach allows for a holistic evaluation of the framework’s effectiveness across various cyber-attack types. The publicly available datasets include UAV attack scenarios and network intrusion data, which cover a wide range of cyber threats. Additionally, we collected data from simulations of different drone operations, including surveillance and delivery missions, under various cyber-attack conditions. The proposed framework demonstrates significant improvements in real-time threat detection for drones, utilizing deep learning and machine learning techniques. The framework was tested using both publicly available datasets and simulations of drone operations under various attack scenarios, achieving high accuracy (95.3%), precision (94.8%), recall (93.7%), and F1-score (94.2%) for CNN-based threat detection. These results highlight the robustness of our approach in enhancing the security and reliability of drone operations. The study contributes to the field of drone cybersecurity by offering a scalable and real-time defense mechanism, supported by a forensic investigation framework. Comparisons with existing techniques highlight significant improvements achieved by our approach. Furthermore, we present case studies that illustrate the practical application of our framework in real-world scenarios, showcasing its capability to handle both cyber-attack and forensic investigation situations effectively. Finally, we address the challenges and limitations encountered during the research, providing insights into potential future work. This paper contributes significantly to the field of drone cybersecurity and forensic investigation, offering a holistic approach that enhances the safety and reliability of drone operations.
Keywords:	Drone Cybersecurity, Forensic Investigation, Deep Learning, Machine Learning, UAV Attack Scenarios, Network Intrusion Data, Threat Detection, Anomaly Detection, Cyber Defense Framework, Digital Forensics, Performance Metrics, Case Studies
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	HYBRID ENSEMBLE ADVANCED MODEL FOR FARMER’S DECISION-MAKING IN CROP MANAGEMENT
Author:	G.ALEXANDAR NARKUNAM, K.KALA, S. SANTHOSH KUMAR
Abstract:	India is the second largest crops producers among all the states. Among the main foods that have been cultivated are wheat and rice. This farming technique aids in increasing crop productivity and facilitating farmers' ability to make the best decisions. High-quality products, disease resistance, and fertilizer responsiveness are the greatest ways to increase crop yield and food quality. The data for this research was gathered using information about the crops, soil, weather, and yield statistics for different crops. The gathered data is sent to the selection process after being carefully pre-processed using the techniques of normalization and intelligent imputation. Using the encoding method and the feature engineering process, the crop is selected based on the input provided as a data set. The LeNet-5 approach is used to determine the optimum crop based on transfer learning, allowing the farmer to choose the crops with the highest yields. With the aid of the hyper tuning procedure, a shallow neural network is used for both the crop analysis and decision-making. This research proposes a hybrid shark smell with Jaya optimize approach for crop tweaking and yield optimization. The suggested hybrid ensemble approach is deployed and continuously integrated to achieve the ideal yield so that farmers can make the right decisions.
Keywords:	Normalization, LENET-5, Shallow Neural Networks, Hybrid Shark Smell, Jaya Optimize.
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	ENHANCING DIABETES CARE VIA ARTIFICIAL INTELLIGENCE
Author:	TURJA BHATTACHARJEE, MOHAMED EL-DOSUKY2, SHERIF KAMEL
Abstract:	Artificial intelligence (AI) has become a potent tool in healthcare with the potential to completely change the way diabetes is treated. This study investigates how AI affects patient outcomes and diabetes treatment. Healthcare providers can extract insightful information from patient data using machine learning, data analytics, and AI-driven wearable devices, resulting in individualized treatment programs and better glycemic control. AI chatbots and virtual assistants improve patient support and engagement, encouraging improved treatment adherence. Despite privacy and ethical issues, AI is effective at cutting healthcare expenses and improving the quality of life for patients is obvious. Healthcare providers can use AI to develop a patient-centered strategy and improve diabetes care by working with researchers and politicians. This paper proposes a smart chatbot for enhancing diabetes care through natural language interactions. The chatbot's architecture uses pattern matching and keyword identification techniques to follow a multi-level interaction procedure. The proposed chatbot system simplifies diabetes diagnosis by using natural language interactions, asking questions based on previous responses through a multi-level diagnostic flow. It employs AIML-based memory techniques and pattern matching to identify keywords at each level, ensuring relevance and coherence in conversation. The system follows a search engine-like flow, using methods like the Sequence Words Deleted (SWD) technique and Triangular Number equation to optimize keyword matching, with Vpath values guiding the diagnostic path. The chatbot enhances patient diagnosis by providing structured, personalized guidance through these techniques.
Keywords:	Artificial Intelligence, Diabetes, Chat-bot, Care
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	COMPOSITE CROSS ATTENTION NETWORK FOR RELIABLE AND ROBUST PLANT NUTRITIONAL DEFICIENCY ANALYSIS: IMAGE ENHANCEMENT AND TRANSFER LEARNING
Author:	MRS. NIKITHA S, DR. PRABHANJAN S
Abstract:	Nutritional deficiency analysis is crucial to enhance agricultural productivity and promote environmental sustainability. This study investigates domain-specific preprocessing techniques to improve the visual features of leaf images. Our proposed image enhancement pipeline uses edge enhancement filters, multiscale Contrast Limited Adaptive Histogram Equalization (CLAHE), and multiscale Retinex to improve the structural, textural, and colour details. "Composite Cross-Attention Network" is proposed to effectively integrate information from both primary and enhanced images. The image features were effectively extracted and integrated using transfer learning and a self-residual cross-attention block. The model achieved promising results with a 5-fold cross-validation strategy on the Mulberry and Rice datasets, achieving accuracies of 92.53% and 97.23%, respectively. It demonstrates superiority over models relying solely on primary images and shows robustness in real-time evaluations under varied environmental conditions. This study underscores the importance of advanced image processing and deep learning integration for optimizing nutritional deficiency analysis in precision agriculture, contributing to sustainable agricultural practices.
Keywords:	CLAHE, Retinex, DenseNet121, MobileNet, Cross-Attention
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	EFFICIENCY OF VIRTUAL REALITY TECHNOLOGIES IN THE DEVELOPMENT OF STRATEGIC THINKING OF FUTURE PROFESSIONALS
Author:	OLENA BULAVINA, DMYTRO KYSLENKO, VITA HMIL-CHUPRINA, SVITLANA ARTEMENKO, LIDIIA CHEREDNYK
Abstract:	This research presents the results of studying the impact of the use of VR in the educational process of students - future specialists, on the development of strategic thinking in them. The purpose of the study is to analyse the influence of virtual reality on the formation of strategic thinking of future professionals. Methods: modelling, pedagogical experiment, expert assessment, survey, methods of mathematical statistics (t-test, Pearson correlation coefficient). For the experiment conduction, Oculus Rift for possible 3D visualisation of professional situations was used, and certain topics of training activities have been developed. The received results showed statistically significant differences in the evaluation of the development of strategic thinking of students, who used virtual reality technologies in the educational process (experimental group – 3.46-3.98 points; control group – 2.67-3.16 points out of maximum 7). Evaluation of the quality of students’ interaction with objects in virtual reality demonstrates that components of interaction such as quality of sensory perception and truthfulness of objects were lower than quality, comfort, and evaluation of task performance. Correlation relationships between some components of evaluation of interaction with virtual reality and components of strategic thinking were strong, besides such components as “truthfulness of objects” and “quality of sensory perception”. The received results have an important meaning for VR use in pedagogical practice for understanding the peculiarities of practical professional activity of future professionals. Further studies/perspectives may be directed at the study of the experience of VR use during classes of different types (including those with different quality of interaction with objects in virtual reality).
Keywords:	Learning Technology, Oculus Rift, Means of Education, Professional Qualities, Immersion Technologies, Development of Strategic Thinking of Future Specialists, Virtual Reality Technology in Professional Education
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	INTEGRATING MULTIMODAL MEDICAL IMAGING DATA FOR ENHANCED BONE CANCER DETECTION: A DEEP LEARNING-BASED FEATURE FUSION APPROACH
Author:	LAKSHMI NAGA JAYAPRADA GAVARRAJU, A. SREENIVASA RAO, RUDRARAJU ANUSHA, DESIDI NARSIMHA REDDY, JYOTHI ANANTULA, , DIVVELA SURENDRA, PRADEEP JANGIR, Dr. SIVA KUMAR PATHURI
Abstract:	Early detection and treatment planning depend on accurate bone cancer detection. Using deep learning-based feature fusion techniques, we propose a novel approach for improving bone cancer detection using multimodal medical imaging data. The method enhances the detection accuracy by combining complementary information from different imaging modalities, including X-ray, MRI, and CT scans. Using a deep fusion architecture, we combine discriminative features from each modality using convolutional neural networks (CNNs). Our results demonstrate the effectiveness of the proposed method in achieving superior detection performance across a diverse dataset of bone cancer patients. A growing number of deep learning models have demonstrated excellent performance on tasks like malignancy rate assessment, grading, segmentation, classification, volume calculation, and detection in primary and metastatic bone tumors using radiological modalities like X-ray, CT, MRI, and SPECT scans along with pathological images. These results point to the possibility of using deep learning to help in bone tumor detection and prognosis prediction. In this paper, we examine the present uses of deep learning-based artificial intelligence in the diagnosis and prognosis prediction of bone cancers, as well as the workflows of these methods in medical imaging. We also go into great detail on the current difficulties in applying deep learning techniques and provide future directions for this developing discipline. To minimize the limitations associated with individual imaging techniques and improve the robustness of bone cancer detection, we combine the strengths of multiple imaging techniques. So, in this article, we proposed a classifier named DTXGB-ResNet50(DEEP TRANSFER XGB-RESNET-50) classifier and compared it with existing classifiers like K-Nearest Neighbors (KNN) and Decision Tee in which the proposed algorithm outperformed when compared with the base classifier’s i.e., 96%.
Keywords:	Bone Cancer, Deep learning, XGB-ResNet-50, AI, Image Processing, KNN, DT.
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	SYSTEMATIC SURVEY ON CREDIT CARD FRAUD TRANSACTION DETECTION TECHNIQUES
Author:	Dr. DV NAGA RAJU, Mr.G VAMSI KRISHNA, Mrs. MADHAVI DEVI LANKA, R. CHANDRA MOHAN, Dr LAKSHMI RAMANI BURRA, Mr BALAJI TATA
Abstract:	MasterCard (CC) plays a critical norm in the current riches. It transforms into a vital piece of the nuclear family, business, and overall activities. While utilizing CCs can offer tremendous benefits whenever utilized circumspectly and securely, huge credit and monetary harm can be brought about by fake action. False Visa exchanges cost firms and purchasers enormous monetary misfortunes consistently, and fraudsters persistently endeavor to track down new innovations and strategies for committing deceitful exchanges. The discovery of false exchanges has turned into a huge component influencing the more prominent use of electronic installment. Consequently, there is a requirement for proficient and compelling methodologies for distinguishing misrepresentation in charge card exchanges. This paper presents Writing review on Visa Extortion Recognition strategies. The principal point is to get Visa exchanges; so individuals can utilize e-banking securely and without any problem. The qualities, whether positive or negative, are referenced for misrepresentation identification strategies. It likewise specifies the as of now utilized cutting edge procedures to counter these assaults and features its limits.
Keywords:	Credit Card, Fraudulent Transactions, E-Banking.
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	THE CLASSIFICATION AND THE PREDICTION OF HEART DISEASE USING THE PROPOSED YOLO TINY ARCHITECTURE
Author:	ELAVARASI.C, DR. M. PRIYA
Abstract:	Among humanity’s the heart disease is one of the most common dangerous disease in the world recently. According to the survey in 2020 somewhere around 4.21 million children’s and adults diagnosed with the heart disease due to the stress and abnormal diet. So to identify and classify this some diagnosing techniques is used to identify and classify very early is done. To identify this manually more time will be taken to find the exact severity of heart disease. So in this paper the artificial intelligence based technique is used to solve /these problems. In this paper the early diagnosis is done using the previous images of heart. Here the GLCM+RCNN is used for the feature extraction so the redundancy will be omitted and the high dimension data will be finalized. Before extracting the feature the preprocessing of the data has to be done. For this the Histogram equalization has been used for the better results. After making as a high dimension data reduction the data reduction is done using the Kernal PCA and this helps in the next step for classifying. For the classification the YOLO Tiny Architecture has been used and this helps in finding the heart disease. By using this method the sensitivity of heart disease is found around 98.9% precision out of 100%, accuracy among 98.1% and recall of 98.7%.
Keywords:	YOLO Tiny Architecture, Histogram Equalization, GLCM+RCNN, Kernal PCA, Classification, Data Reduction, Feature Extraction, High Dimension Reduction, Preprocessing, Heart Disease.
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	CONTOUR-BASED DIACRITICS DETECTION FOR ENHANCED ARABIC TEXT IMAGE PROCESSING
Author:	TARIK ABDEL-KAREEM ABUAIN
Abstract:	Using diacritics in words or letters are not merely additional or optional elements of the language, which make them essential components in many scripts. Altering certain diacritics can significantly affect both the syntax and semantics, potentially changing the word's meaning into a completely different one. Detecting Arabic diacritics accurately remains a critical yet challenging task within the world of document image processing, especially in languages where diacritics significantly impact meaning and pronunciation. However, much research addresses the primary objects (letters) in text and neglects the secondary objects (diacritics) and considers them as noise. Therefore, this research presents a contour-based methodology for diacritics detection (segmentation) to improve the quality and efficiency of image-processing techniques applied to Arabic handwritten texts by exploiting the features that can be extracted from the detected diacritics in both machine written and handwritten text images. The proposed method involves converting text images to grayscale, applying adaptive thresholding to produce binary text images, and employing contour tracing method to isolate diacritic regions. This method was tested on a self-created dataset of over 50 Arabic machine printed and handwritten text images. The results were very promising, where the accuracy rate of the proposed method on the Arabic machine printed text images achieved between 0.97% to 0.98%, while for the Arabic handwritten text images the proposed method achieved between 0.92% to 0.99% accuracy rate depending on the used threshold values per area in the text.
Keywords:	Text Image Segmentation, Diacritics Detection, Contour-Based Methodology, Image Processing
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	A NEW HYBRID DEEP LEARNING MODEL FOR DIABETIC RETINOPATHY DETECTION
Author:	RACHNA KUMARI, SANJEEV KUMAR, SUNILA GODARA
Abstract:	The progressive eye disease known as diabetic retinopathy continues to be the leading cause of blindness worldwide. Effective treatment and prevention of vision loss require prompt and accurate DR detection. Profound learning procedures have shown extraordinary commitment in clinical picture examination, and in this paper, we propose a hybrid model that joins the qualities of convolutional brain organizations (CNNs) and repetitive brain organizations (RNNs) to further develop DR discovery exactness. The proposed crossover profound learning model involves three principal stages. A pre-handling, first and foremost, step is applied to upgrade the quality and differentiation of fundus pictures, in this manner working on the model's capacity to remove basic highlights. After that, a Residual CNN is used to extract features from the images that have already been processed. Residual CNNs are adroit at catching various leveled highlights, and this stage empowers the model to successfully gain discriminative elements from the information pictures. The subsequent stage includes incorporating RNNs into the model. RNNs are ideal for analysing sequential patterns in medical images because they are well-suited to handling sequential data and capturing temporal dependencies. The model's ability to extract temporal information from fundus image sequences thanks to the inclusion of RNNs improves its ability to identify early DR progression signs. The third and last stage centers around the characterization task, where a completely associated brain network is utilized to decipher the highlights separated by the past stages and order the pictures into various DR seriousness levels. The hybrid model's architecture facilitates the fusion of spatial and temporal information, resulting in a more comprehensive and accurate DR diagnosis.
Keywords:	Diabetic Retinopathy, Deep Learning, Hybrid Model, Detection, Retinal Images.
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	A NOVEL MODEL BASED ON DEEP TRANSFER LEARNING FOR DETECTING MALICIOUS JAVASCRIPT CODE
Author:	XUAN DAU HOANG, THI THU TRANG NINH, HOANG DUY PHAM
Abstract:	The escalating prevalence of cyber threats and malware attacks across multiple platforms in recent years has highlighted the need for automated machine learning defense mechanisms. Numerous studies have focused on leveraging deep learning to identify malicious JavaScript code, showing promise in improving cybersecurity measures. Moreover, advancements in large language models (LLM), particularly generative pre-trained transformer-based models like GPT-2/3, have also created opportunities for more effective cyber threat prevention. Overall, these developments point to the significant potential of deep learning techniques for the efficient training of models and effective detection of threats within JavaScript code. This paper proposes a novel deep transfer learning-based model for detecting malicious JavaScript code using CodeBERT to improve the detection performance and minimize manual data engineering tasks. Since CodeBERT can be fine-tuned to adapt to different downstream tasks, we formulate different approaches based on CodeBERT to explore possible scenarios. We then evaluate our approaches on various datasets, and compare the performance of our models with previous researches, as well as baseline models, including both deep learning and traditional machine learning methods. Experimental results confirm that our CodeBERT-based model can detect malicious JavaScript code efficiently on various experimental datasets with the F1-score of 99.3%, which is better or comparable with results of the state-of-the-art proposals.
Keywords:	XSS detection model, malicious JavaScript detection, XSS detection based on deep transfer learning, CodeBERT-based XSS detection model
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	ENHANCING IOT SECURITY THROUGH MALWARE CLASSIFICATION WITH HYBRID DEEP LEARNING MODEL GENERATIVE ADVERSARIAL NETWORK AND DEEP BELIEF NETWORK (HCAGAN_DBN)
Author:	V.S. JEYALAKSHMI, KRISHNAN NALLAPERUMAL
Abstract:	IoT has gained popularity as a result of the advancement and promise of smart technologies. The necessity of IoT technology has been accompanied by an increase in security issues about IoT devices, apps and infrastructure. Due to the wide range of abilities of IoT devices, dynamic and the constantly changing environment, enhanced system security measures are challenging and it is risky to only implement the most fundamental security requirements. The computer system is significantly at risk from malicious software (Malware). Finding malicious intent in a program is a crucial responsibility for security purpose. Here, novel method is proposed to detect significantly the previously unidentified threats in a cyber security land space. A new hybrid model HCAGAN-DBN is developed to classify the malware family efficiently. GAN architecture has generator and discriminator, the re-sampled output data is transformed into DBN for malware family classification in a zero short learning. Generative Adversarial Network and Deep Belief Network based intrusion detection model is proposed in this paper for Malware classification in IoT environment. The proposed model trained for one-dimensional images which learn and analyzes the characteristics of the complicated patterns of the Byte files and the Assembly files. Experiments were carried out with the Microsoft malware Challenge Dataset (MMCD). The outcomes of the evaluation show that with an average accuracy of 99.83% our HCAGAN_DBN Classifier performs better than traditional state-of-the-art works.
Keywords:	Internet of Things, Generative Adversarial Network, Cyber Security, Malware Analysis, Deep Belief Network
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text

Title:	REMOTE MONITORING OF THE DRIVER’S HEART RATE, BREATHING RATE AND FATIGUE SIGNS USING A HYBRID APPROACH ON DIFFERENT EMBEDDED ARCHITECTURES
Author:	HODA EL BOUSSAKI, RACHID LATIF, AMINE SADDIK
Abstract:	In this work, a hybrid approach to remote monitoring of the heart rate, the breathing rate and the driver’s fatigue signs is presented. The outcomes have been applied across various architectures, including a Desktop, the Odroid XU4, the Jetson Nano and the Odroid H2. The proposed algorithm uses the Dlib library for face detection, facial landmarks and the extraction of the region of interest. It also utilizes remote plethysmography, which is a video-based method allowing non-contact monitoring of blood volume fluctuations by identifying changes in pixel intensity on the skin, for the estimation of the heart rate and the respiratory rate. The efficiency of Dlib’s face detection capabilities is assessed in comparison to other algorithms, including those optimized for GPU implementation.
Keywords:	Heart rate, Breathing rate, Fatigue, Driver, Non-contact
Source:	Journal of Theoretical and Applied Information Technology 30^th September 2024 -- Vol. 102. No. 18-- 2024
Full Text