ABC-GAN: An Attention-Based Conditional GAN with DTW Loss for ECG Signal Generation to Address Class Imbalance

perRICESTJournal of Information Systems and Telecommunication (JIST) 2322-14372345-27732026-0914547998articleABC-GAN: An Attention-Based Conditional GAN with DTW Loss for ECG Signal Generation to Address Class ImbalanceParmida Behainbehain.parmida@gmail.com1Noushin Riahinriahi@alzahra.ac.ir2The imbalance between normal and pathological cases in Electrocardiogram (ECG) datasets significantly degrades the performance of automated deep learning-based diagnostic systems, particularly for minority classes. This paper introduces ABC-GAN, a novel Attention-Based Conditional Generative Adversarial Network designed to mitigate this critical data imbalance by generating high-fidelity, class-specific synthetic ECG signals. Our proposed model incorporates two key innovations: an attention mechanism within the generator to focus on critical morphological features of the ECG waveform, and the integration of Dynamic Time Warping (DTW) as a distance metric in the generator's loss function to better preserve essential temporal dynamics. Trained and thoroughly evaluated on the MIT-BIH Arrhythmia dataset across seven different heartbeat classes, ABC-GAN successfully generates highly realistic signals that closely match the original data's distribution. When used to augment the training set, these synthetic signals significantly enhance classifier performance and generalization. A downstream classifier achieved an accuracy of 95%, representing a substantial improvement over the baseline trained on the raw imbalanced data and clearly outperforming both standard oversampling techniques and a vanilla Conditional GAN (cGAN). The overall findings demonstrate that ABC-GAN is a powerful and effective tool for data augmentation, fully capable of improving diagnostic accuracy in cardiac research and practical clinical applications.http://jist.ir/Article/51397ECG Signal Synthesis Generative Adversarial Network (GAN) Attention Mechanism Conditional GAN Dynamic Time Warping (DTW) Class Imbalance Data AugmentationperRICESTJournal of Information Systems and Telecommunication (JIST) 2322-14372345-27732026-09145499116articleOptimization of Hyperparameters of BERT Model in Sentiment Analysis Using Genetic AlgorithmShahla Sadeghanish.sadeghani1989@gmail.com1Mohammad jafar Tarokhmjtarokh@kntu.ac.ir2Mohammad Ali Afshar Kazemidr.mafshar@gmail.com3Sentiment analysis, a vital branch of natural language processing (NLP), is progressing rapidly, and advanced models such as BERT are used to improve the accuracy of such analyses. However, tuning BERT’s hyperparameters remains a major challenge, directly influencing its performance. The methods used in previous research to optimize hyperparameters often face limitations in accuracy and efficiency. This study addresses an important research knowledge gap by using genetic algorithm (GA) to present a new approach to optimize the hyperparameters of BERT model in sentiment analysis with the main goal to get better results. In this regard, this research is designed analytically and experimentally. Data were collected through Twitter API and preprocessed using NLP techniques including noise removal, tokenization, and text normalization before being analyzed.  The BERT model’s hyperparameter optimization was performed using GA and the optimized models were evaluated based on criteria such as accuracy, precision, recall and F-1 score.  The proposed algorithm was compared with other common methods such as vanilla BERT, Long Short-Term Memory (LSTM) and convolutional neural network (CNN). The results showed that the proposed model has significantly better performance than other methods, achieving 98.1% accuracy, 98.2% precision, 98.1% recall, 98.15% F-1 score, and an area under the curve (AUC) of 98.5%. In comparison, vanilla BERT model, LSTM and CNN scored 92.8%, 91.3% and 90.5% in accuracy, respectively. These results indicate that the use of GA in optimizing the hyperparameters of BERT model can effectively increase the accuracy and efficiency of sentiment analysis. This approach not only has applicability in various fields of text data analysis, but also can be used as an effective solution for future research in optimizing deep learning models.http://jist.ir/Article/51529Sentiment Analysis BERT Genetic Algorithm Hyperparameter OptimizationperRICESTJournal of Information Systems and Telecommunication (JIST) 2322-14372345-27732026-091454117128articleImproving the Accuracy of Motor Imagery in BCI for the Movement of Artificial Prostheses used for Disabled PeopleIsaac Jahanbakhshiisaac.jahanbakhshi@yahoo.com1shaghayegh Niavaraniniavarani.sh@gmail.com2Fatemeh Shahbazifatimashahbazi800@gmail.com3Farahnaz Abdifarahnaz.abdi65@gmail.com4 A brain-computer interface (BCI) allows users to communicate directly with an external device, such as a computer, using brain signals. These systems involve signal acquisition, temporal and spatial filtering, feature engineering, and classification before transmitting the control signal to an external device. Brain-computer interfaces based on motor imagery (MI-BCI) hold significant potential for applications in motor enhancement and rehabilitation. However, the control capabilities of MI-BCI can vary among individuals. In this article, we present a motor imagery model that leverages the power of deep learning for feature extraction and optimization methods for feature selection, based on individuals' electroencephalogram (EEG) signals. For this purpose, we employed the convolutional neural network (CNN), the golden eagle optimization (GEO) method, and three hybrid classification models to achieve high accuracy in classifying EEG signals. The innovative classification method presented in the third phase of the proposed method has high flexibility and significant accuracy, which is presented for the first time. This method can be used in all matters of prediction and detection of phenomena to increase accuracy and high scalability. The experimental results demonstrate that the proposed model significantly outperforms baseline approaches across all key performance indicators. In terms of precision, the method achieves an improvement of approximately 12–18% over standard CNN and nearly 20% compared to conventional classifiers. For recall, it yields 10–15% higher performance than CNN and 18–22% higher than classical machine learning methods. Finally, the F-measure results indicate a 12–15% improvement over CNN and about 20% over traditional classifiers. These consistent enhancements confirm the robustness and scalability of the proposed approach for motor imagery-based EEG classification and highlight its potential for practical applications in rehabilitation and prosthetic control systems.  http://jist.ir/Article/48179Brain-Computer Interface Motor Imagery CNN GEOperRICESTJournal of Information Systems and Telecommunication (JIST) 2322-14372345-27732026-091454129141articleMachine Learning Ensemble Model for Predicting Adoption of Metaverse in Higher Education Using PSO Algorithm for Setting Model’s HyperparametersAtaollah Abtahia-abtahi@srbiau.ac.ir1zahra Afjeizafjeh@gmail.com2Ebrahim Nazari Farokhie60_itmgtn@yahoo.com3In recent years, the metaverse has rapidly gained prominence as an emerging virtual platform with multidimensional and interactive features. Its potential applications, particularly in higher education, have drawn increasing attention. However, the acceptance of the metaverse in educational contexts depends on a range of factors that remain insufficiently explored. This study aims to address gaps in previous research, where certain critical factors—such as awareness, user experience, and technological challenges affecting faculty and students in Iran—have been overlooked. The goal of this research is to develop and evaluate an ensemble machine learning model to predict metaverse adoption in Iranian higher education institutions. Drawing on questionnaire data and guided by metaheuristic parameter tuning, the model seeks to identify and forecast factors influencing the uptake of this technology. This descriptive-analytical study collected data from 730 respondents, including university students and faculty members, who answered a questionnaire on metaverse acceptance. Four machine learning models—Random Forest, XGBoost, and Gradient Boosting—were employed. Their parameters were optimized using the PSO (Particle Swarm Optimization) metaheuristic algorithm. Performance metrics included Accuracy, Recall, Precision, and the F1-Score. The results showed that ensemble machine learning models, enhanced by metaheuristic parameter tuning, accurately predicted metaverse adoption. The ensemble model achieved an outstanding accuracy of 95%. Moreover, variables such as technological accessibility, familiarity with the metaverse, and institutional support emerged as key determinants. This research demonstrates that ensemble machine learning models, combined with metaheuristic parameter optimization, can serve as effective tools for forecasting metaverse acceptance in higher education. The findings can help educational administrators devise more effective strategies for implementing and fostering the use of the metaverse, ultimately improving the integration of this technology into academic environments.http://jist.ir/Article/48873Metaverse Adoption Higher Education Machine Learning Metaheuristic SettingperRICESTJournal of Information Systems and Telecommunication (JIST) 2322-14372345-27732026-091454142152articleA Hybrid Deep Neural Network for Arabic Fake News Classification based on Temporal CNN and BiLSTM with Attention Azhar A Hadiazhar@uobabylon.edu.iq1Abbas F Hommadiabbas.hommadi@uobabylon.edu.iq2Hussein A Ismaelhusseinyessari@uobabylon.edu.iq3The proliferation of fake news on social media poses a significant threat to societal harmony, especially in languages like Arabic, which is distinguished by its complexity. Thus, Artificial intelligence techniques are necessary to mitigate the impact of misinformation. Many researchers have conducted this challenge by introducing machine learning and deep learning approaches. This study presents a new hybrid deep learning network for enhancing the classification of Arabic fake news. This network combines three mechanisms: Temporal Convolutional Networks, Bidirectional Long Short-Term Memory, and Attention Mechanism. This mixture model produces a strong feature extraction process with temporal awareness. Moreover, the attention layer enables the model to focus only on relevant features and ignore irrelevant ones to concentrate on salient features. Also, several stages of pre-processing Arabic text, representing words using a pre-trained word embedding model (Glove, Ara2Vec), and extracting features through advanced layers (BiLSTM, TCN, and Attention). Extensive experimentation on large-scale and well-known Arabic fake news datasets (AFND and AraNews) demonstrates the efficacy of the proposed model. The hybrid model shows superior performance compared to baseline and previous state-of-the-art studies by achieving an accuracy of 88% and 95% on the AFND and AraNews datasets, respectively. Our results show that the suggested model can be used as a powerful technique to restrain the widespread online misinformation in the Arabic language.http://jist.ir/Article/49506Arabic Fake News Temporal Convolution Network BiLSTM Attention Mechanism Deep LearningperRICESTJournal of Information Systems and Telecommunication (JIST) 2322-14372345-27732026-091454153169articleSmartening and Digital Transformation Plan in the Industry based on Best PracticesMehdi Azadimotlaghm.azadim@pgu.ac.ir1Reza Sharafdinisharafdini@pgu.ac.ir2Zahra Salimizahra.2002.salimi@gmail.com3Melika khosravim.khsrv1998@gmail.com4Yekta Farzadkiayektafarzadkia04@gmail.com5Due to smart production, Industry 4.0 has become an essential paradigm in industry. This paradigm is looking for smartening and digital transformation in industries where its goal is not only technology upgrade but also integration of advanced technology such as Artificial Intelligence, Internet of Things, Cloud computing, Digital twins and simulation software, Big Data Processing, Blockchain, etc. in production processes for increasing efficiency, flexibility, and productivity in various sections of production. Research shows that smartening and digital transformation are essential to creating a competitive advantage in global markets and leadership in the coming decades. Neglecting this change of approach can lead to losing customers and sales markets. Unfortunately, despite its advantages, implementing the smart industry faces significant challenges that can hinder its achievement of goals. So, whether at the national or corporate level, benefiting from a cohesive program for smartening and digital transformation in any industry is essential. In this paper, based on research in more than 20 leading countries in the field of smart industry, a strategic plan for smartening and digital transformation in industries has been presented. By using this strategic plan, it is possible to overcome the challenges and reach the smart industry in various fields.http://jist.ir/Article/50887Smartening Digital Transformation Industry 4.0 Executive PlanperRICESTJournal of Information Systems and Telecommunication (JIST) 2322-14372345-27732026-091454170181articleA Spectral-Domain Approach for Early Blockage Detection in Sub-Terahertz Wireless NetworksNeravati Nagaraja Kumarrajuneravati@gmail.com1Anil Kumaranidecs@gmail.com2Subba Raju MPsubbaraju.56@gmail.com3Kalyani Kapulakalyani.kapula@gmail.com4Surya Kala Nagireddikala.nagireddi@gmail.com5Yarrapragada Rao K. S. S. subbaraoyarrapragada@gmail.com6The paper presents the problem of human body blockage in sub-terahertz and millimeter wave wireless systems. These communication systems are operated at very high frequencies. The generated high frequency signals are very sensitive to obstacles like the human body. A blockage event abruptly reduces the received signal power. This causes a loss of signal connection between transmitter and receiver. To avoid this problem, the network detects a blockage before it actually happens. Many existing solutions use machine learning models in the time domain. These methods are complex. Long training times are required. The signal shows clear oscillations just before a blockage happens. These oscillations are weak in the time domain. The paper uses the short-time Fourier transform to extract spectral features from the received signal. The difference between normal conditions and pre-blockage conditions becomes very large. In some cases, the gap reaches two orders of magnitude. Based on this observation, a threshold-based proactive blockage detection algorithm is designed. The algorithm is simple and does not rely on machine learning. In order to solve this, MATLAB simulation environment is assumed with parameter values and data collected at 156~GHz in an indoor environment. The performance is measured using blockage detection probability, mean time to blockage and false alarm rate. The paper explains the spectral analysis offers a simple and effective way to enable proactive blockage detection for future sub-terahertz wireless networks.http://jist.ir/Article/49037Sub-Terahertz Communication Human Body Blockage Proactive Blockage Detection Short-Time Fourier Transform Spectral-Domain Analysis Threshold-Based Detection Algorithm