Articles

Access the latest knowledge in applied science, electrical engineering, computer science and information technology, education, and health.

Advanced Search
Filter Icon

Filters article

Years

FAQ Arrow
Single Year
Range Year
0
0

Source Title

FAQ Arrow

Authors

FAQ Arrow

31,377 Article Results

Design of intelligent embedded system for personal protective equipment detection and face recognition access control

10.11591/ijai.v15.i4.pp3176-3188
Mariam Mesfer , Aoosh Matar , Reem Saif , Alyazy Saleh , Irfan Ahmed , Moath Awawdeh , Anees Bashir
This paper presents an artificial intelligence (AI)-powered automated access control system that aims to reduce delays and improve safety. The primary problem addressed is effective monitoring of compliance with personal protective equipment (PPE) and secure access control for personnel entering sites. This study represents the design and development of an access control system that includes accurate detection of essential PPE items (e.g., safety helmets, gloves, goggles, and gas detectors), integration of facial recognition for identity verification, real-time monitoring of video feeds, and an intuitive user interface for security personnel to manage access and compliance efficiently. The software part uses you only look once (YOLO) version 8 for real-time object detection, classification, and drawing the bounding boxes around the detected object in a single forward pass. The hardware platform consists of NVIDIA Jetson AGX Orin 64 GB as an edge computing device. The developed AI-based embedded system is tested and validated with real-world scenarios and achieved a mean average precision (mAP) of 98.4% for PPE detection and 99.38% accuracy for face recognition.
Volume: 15
Issue: 4
Page: 3176-3188
Publish at: 2026-08-01

Energy loss prediction using least absolute shrinkage and selection operator regression and SHAP explainability

10.11591/ijai.v15.i4.pp3103-3119
Nur Diana Izzani Masdzarif , Siti Azirah Asmai , Yogan Jaya Kumar , Muhammad Hafidz Fazli Md Fauadi
Technical energy loss estimation in power distribution systems is essential for improving operational efficiency and cost-effectiveness. However, distribution-level datasets are often failed to cope with the nonlinear behavior and sparse, low-resolution data typical in modern grid environments. This study proposes an interpretable artificial intelligence (AI) framework based on least absolute shrinkage and selection operator (LASSO) regression integrated with Shapley additive explanations (SHAP) to estimate and explain technical energy losses at the feeder level. An exploratory multicollinearity assessment using correlation analysis and variance inflation factor (VIF) revealed severe redundancy among operational variables, justifying the adoption of L1-regularized regression. Hyperparameter tuning via LassoCV identified an optimal regularization parameter, resulting in strong predictive performance. Comparative evaluation with nonlinear models, including random forest and gradient boosting, demonstrated that LASSO achieves competitive or superior generalization performance while preserving interpretability. Feature importance analysis and SHAP-based explanations confirmed that operational loading variables particularly infeed energy, load factor, and maximum demand are the dominant drivers of technical losses. SHAP dependence and interaction analyses further revealed context-dependent behavior among correlated predictors, enriching interpretability beyond coefficient-based rankings. The results demonstrate that regularized linear modeling, when combined with explainable AI techniques, provides a robust, transparent, and practically deployable solution for technical loss estimation in distribution networks.
Volume: 15
Issue: 4
Page: 3103-3119
Publish at: 2026-08-01

Utilizing Poisson modeling and advanced machine learning techniques for enhanced detection of slowloris attacks

10.11591/ijai.v15.i4.pp3154-3163
Prathima Mabel John , Mamatha Ram , Vani Kurugod Ashwathanarayana Reddy , Santhosh Babu , Rama Mohan Babu Kasturi Nagappasetty
This paper presents a secure framework for detecting and mitigating slowloris attacks in software defined networks (SDN) using Poisson arrival modeling to generate attack traffic. Slowloris attacks imitate legitimate traffic patterns, making them difficult to identify. The framework includes three modules: a data generation module, a feature selection module based on the blended statistical and information gain (BSIG) approach, and a traffic classification module using machine learning and deep learning algorithms. Poisson arrival modeling captures the stochastic nature of attack traffic by introducing variability in packet inter-arrival times, thereby simulating realistic network conditions. Monte Carlo simulations establish anomaly thresholds, while the BSIG method identifies key features, such as inter-arrival variability and CPU utilization. Support vector machine (SVM), k-nearest neighbors (KNN), and random forest (RF) classifiers achieve high precision in distinguishing legitimate traffic from attack traffic. The integration of Poisson modeling with advanced analytical methods improves detection capability while maintaining scalability for similar stealth attacks. The proposed methodology provides high detection accuracy and adapts to dynamic network environments, enhancing web server defenses against slowloris attacks in real-world scenarios. The framework also reduces false positives, supports efficient resource utilization, and strengthens proactive security management in modern SDN infrastructures under varying conditions.
Volume: 15
Issue: 4
Page: 3154-3163
Publish at: 2026-08-01

Enhancing early detection of autism spectrum disorder through ensemble-based machine learning classifiers

10.11591/ijai.v15.i4.pp3732-3744
Shabeena Lylath , Laxmi B. Rananavare
Autism spectrum disorder (ASD) is a developmental disability characterized by significant social, communication, and behavioral challenges. Machine learning is a practical approach for autism detection. The proposed ensemble-based machine learning classifier methodology presented in this study seeks to revolutionize the diagnosis of ASD by harnessing the collective power of various machine learning algorithms. This ensemble approach is designed to enhance diagnostic precision, mitigate the subjectivity associated with traditional diagnostic methods, and accelerate the detection process. This methodology addresses the urgent need for early and accurate ASD identification, enabling timely interventions. Leveraging complex data analysis, it offers deeper diagnostic insights, facilitating informed clinical decisions and advancing ASD research. The methodology's accessibility across healthcare settings marks a significant step forward in making early ASD detection more universally available, showcasing the transformative potential of machine learning in healthcare. In deploying the “ensemble-based machine learning classifier” for ASD diagnosis, this study utilizes an extensive dataset comprising behavioral and medical profiles from diverse demographics, including toddlers, children, adolescents, and adults with ASD. Upon the preliminary analysis, the dataset enables the methodology to learn from a wide array of ASD manifestations, ensuring its robustness and applicability across different age groups and severity levels.
Volume: 15
Issue: 4
Page: 3732-3744
Publish at: 2026-08-01

Multi-stage hybrid YOLO-driven and MobileNetV2-CNN variants for robust fish freshness classification

10.11591/ijai.v15.i4.pp3660-3671
Raseeda Hamzah , Rosniza Roslan , Amni Munira Khidir , Lala Septem Riza
This study presents a multi-stage hybrid representation learning framework for robust fish freshness classification to address the critical challenge of reliable quality assessment in unpreserved food supply chains. The proposed pipeline operates in two stages: stage-1 employs you only look once (YOLO)v8n as feature-gated detector to validate inputs and eliminate non-fish images, while stage-2 leverages transfer-learned MobileNetV2 variants enhanced with convolutional neural network (CNN) layers, fine-tuning, adaptive learning rate schedulers, and expanded fully connected layers for hierarchical classification into three freshness classes i.e., highly fresh, fresh, and not fresh. The framework has been trained and evaluated on two curated datasets comprising 4,500 fish and non-fish images and 11,111 freshness-labeled including augmented dataset to increase diversity. The experimental results showed significant improvements. The best performance model transfer learning (TL)-MobileNetV2 + CNN + fine-tuning achieved 98.07% training accuracy and 67.72% validation accuracy on 80:20 split, and training accuracy of 97.57%, validation accuracy of 97.21% through 10-fold cross-validation. The comparative benchmarking confirmed that dual-stage design outperformed baseline MobileNetV2 and YOLOv5s models across precision, recall, and F1-score. The findings highlighted significant value of integrating detection-driven validation with transfer-learning classification, and propose new benchmark for intelligent freshness monitoring. For future work, this study aims to explore attention-based models, data integration, and species diversity.
Volume: 15
Issue: 4
Page: 3660-3671
Publish at: 2026-08-01

Deep learning for categorizing microsatellite stability in colorectal cancer

10.11591/ijai.v15.i4.pp3761-3769
Sofyan El Idrissi , Yassine Drider , Ikram Ben Abdel Ouahab , Mohammed Bouhorma , Fatiha El Ouaai
Cancer remains a significant global health challenge, with its incidence rising steadily in recent decades. In colorectal cancer (CRC), microsatellite instability (MSI), and microsatellite stability (MSS) are important biomarkers that influence treatment decisions and patient outcomes. Accurate MSI classification is critical but traditional methods can be costly and time-consuming. This study explores the potential of deep learning to classify MSI and MSS in CRC. A large dataset of CRC patients with confirmed MSI and MSS status was utilized, obtained through standard testing images. Deep learning models were applied to histopathological images, analyzing tissue features from digital slides. Convolutional neural network (CNN) and residual network (ResNet)-18 models demonstrated high accuracy in distinguishing between MSI and MSS CRCs. The best-performing model, which integrated genomic and histopathological data, achieved an area under the curve (AUC) receiver operating characteristic (ROC) of 0.85, indicating strong discrimination capability. The findings suggest that deep learning could be a valuable tool for clinical decision-making and personalized medicine in CRC.
Volume: 15
Issue: 4
Page: 3761-3769
Publish at: 2026-08-01

Bridging gaps in health artificial intelligence: challenges in MDPI research articles

10.11591/ijai.v15.i4.pp3053-3067
Irwan Bastian , Aqilla Rahman Musyaffa , Lukman Nulhakim , Novia Putri Bahirah , Dewi Agushinta R.
Technological advancements in artificial intelligence (AI) have transformed healthcare by improving early disease detection, personalized treatment, predictive analytics, and clinical decision support systems. However, AI adoption in healthcare faces critical challenges, including data privacy concerns, algorithmic bias, regulatory barriers, usability issues, and system interoperability. Addressing these issues requires standardized regulations, ethical frameworks, and interdisciplinary collaboration to ensure responsible AI integration. This study systematically analyzes research trends in health AI over the past six years through a systematic literature review (SLR) and a bibliometric analysis using VOSviewer. The review focuses on Multidisciplinary Digital Publishing Institute (MDPI) journal articles to identify key contributors, emerging trends, and research gaps in AI-driven healthcare. Findings highlight dominant research areas, including machine learning for diagnosis, AI-driven hospital management, and predictive analytics, while exposing persistent challenges such as a lack of standardized AI models, ethical concerns, and accessibility disparities. By mapping the research landscape, this study provides evidence-based insights and recommendations to address AI adoption barriers, improve transparency, and guide future research in healthcare AI. The results contribute to developing a more equitable, efficient, and trustworthy AI-driven healthcare system.
Volume: 15
Issue: 4
Page: 3053-3067
Publish at: 2026-08-01

Determining student scaffolding levels in geometry problem-solving: a fuzzy inference system using Mamdani method

10.11591/ijai.v15.i4.pp3286-3298
Yuniar Ika Putri Pranyata , Susiswo Susiswo , Tjang Daniel Chandra
This study aims to adopt a fuzzy logic inference system using Mamdani method to determine the appropriate scaffolding level for students based on errors in solving geometry problems. The fuzzy inference system (FIS) assessed the students' understanding and provided scaffolding to address specific learning needs by analyzing common mistakes in geometry problem-solving. This method optimized the educational process by offering personalized support, enhancing students' problem-solving skills, and reducing error rates. The data processing criteria using the FIS involved analyzing the scores of mathematics education students at a private university in Malang when solving geometry problems, based on Pólya's stages. Mamdani was used to provide recommendations based on students' cognitive data while solving geometry problems, in line with scaffolding components. The results showed that the system personalized scaffolding levels for students working on geometry problems. This contributed to the field of educational technology by presenting a new approach to adaptive learning and emphasized the importance of personalized pedagogical support. The approach was particularly beneficial for geometry teachers who provided individualized scaffolding based on students' errors, contributing to improved learning outcomes.
Volume: 15
Issue: 4
Page: 3286-3298
Publish at: 2026-08-01

Hybrid deep learning-based congestion prediction for intelligent traffic management in 5G/6G networks

10.11591/ijeecs.v43.i2.pp472-484
T. Anuradha , Giribabu Sadineni , Laxmi Pamulaparthy , M. L. M. Prasad , N. Vijay , Rajamahendravarapu Lakshmi Durga
Future-generation communication networks, including millimeter-wave local area networks, broadband wireless access systems, and emerging fifth- and sixth-generation (5G/6G) networks, require intelligent traffic management to satisfy stringent quality of service (QoS) requirements, including ultra-low latency, high reliability, and massive device connectivity. As network traffic becomes increasingly dynamic and heterogeneous, accurate congestion prediction is essential for preventing resource overloading, maintaining network slicing performance, and ensuring efficient resource utilization. This paper proposes a hybrid deep learning (DL)-based congestion prediction model that combines long short-term memory (LSTM) and support vector machine (SVM) techniques to capture temporal traffic characteristics while improving prediction accuracy. The proposed framework was evaluated through a one-week simulation involving heterogeneous devices operating under varying network conditions to assess its robustness and generalization capability. Experimental results demonstrate that the proposed model achieved an overall prediction accuracy of 93.23%, while also exhibiting strong performance in terms of specificity, recall, F-score, and computational efficiency. By accurately predicting network congestion before service degradation occurs, the proposed framework enables proactive traffic management and adaptive resource allocation. These findings demonstrate that the hybrid LSTM–SVM model provides an effective, reliable, and scalable solution for intelligent congestion prediction in next-generation 5G/6G communication networks.
Volume: 43
Issue: 2
Page: 472-484
Publish at: 2026-08-01

Advanced personal bankruptcy prediction using tree-based deep learning models

10.11591/ijeecs.v43.i2.pp640-650
Nhat Nguyen Minh , Duy Ngo Hoang Khanh
Due to the unstable economic conditions, worsened by the post-COVID-19 environment and persistent foreign wars in 2024, financial institutions have growing difficulties in accurately predicting customer default probability. This study examines the use of sophisticated tree-based deep learning and deep neural network models for forecasting personal bankruptcy. This research utilises a dataset of roughly 9,800 individuals from Vietnamese financial institutions, spanning from 2012 to 2022, to evaluate the efficacy of models including neural decision tree, deep forest, tabular convolutional neural networks (TBCNN), and neural oblivious decision ensembles (NODE). The results demonstrate that the Deep Forest model far surpasses its competitors, providing nearly flawless predicted accuracy and enhanced interpretability. The findings highlight the efficacy of tree-based deep learning and deep neural network models as effective instruments for financial risk management, especially in volatile and unpredictable economic environments.
Volume: 43
Issue: 2
Page: 640-650
Publish at: 2026-08-01

The next era of electrification: engineering adaptive energy infrastructures for a decarbonized society

10.11591/ijeecs.v43.i2.pp355-362
Tole Sutikno
A new era of electrification is reshaping electrical engineering, with adaptive energy infrastructures becoming critical foundations for achieving resilient, low-carbon, and sustainable energy systems. Rapid advances in renewable energy integration, power electronics, distributed energy resources, battery energy storage, electrified transportation, and intelligent energy management are transforming conventional power systems into flexible, interconnected, and resilient infrastructures. This editorial discusses the emerging paradigm of adaptive electrification, in which future electrical networks are expected to dynamically coordinate generation, energy storage, conversion, transmission, distribution, and consumption across increasingly decentralized environments. Beyond traditional objectives of efficiency and reliability, next-generation electrical infrastructures must address growing electricity demand, carbon neutrality, power quality, climate resilience, cyber-physical security, and energy accessibility. The editorial further highlights several promising research directions, including grid-forming power electronics, adaptive microgrids, intelligent battery management, solid-state energy conversion, digitalized power infrastructures, resilient hybrid AC/DC systems, and coordinated human–AI decision-making. Collectively, these technological advances position electrical engineering at the forefront of the global energy transition, emphasizing that future electrification requires not only cleaner energy sources but also adaptive, intelligent, and sustainable infrastructures capable of continuously evolving to meet societal, environmental, and industrial challenges while supporting reliable and equitable access to electricity.
Volume: 43
Issue: 2
Page: 355-362
Publish at: 2026-08-01

Improved sizing of a damped passive filter for harmonic attenuation using a teaching-learning-based optimization algorithm for an industrial site

10.11591/ijeecs.v43.i2.pp363-383
Brahimi Omar , Rabah Djekidel , Hadjadj Abdechafik , Sid Ahmed Bessedik
Based on measurements of current harmonics that adversely affect power quality in an industrial setting, and to ensure compliance with permissible levels according to IEEE 519, this paper aims to compare the effectiveness of conventional and optimized sizing of damped passive filter components to reduce total harmonic distortion (THD) and improve the power factor (PF). The optimization of passive filter element values relies on a powerful metaheuristic algorithm called teaching-learning-based optimization (TLBO). Simulation results show that the THD of current and voltage reach values of 17.34% and 8.61%, respectively, and decrease to very low values after implementation of the damped passive filter in the optimal case using the TLBO algorithm. The results also demonstrated a proportional relationship between the harmonic currents in the load and the overall increase in losses in the transformer. Optimizing the damped passive filter reduced harmonic distortion, improved the PF, and limited energy losses, thus confirming the superior ability of this algorithm to identify a very high efficiency solution.
Volume: 43
Issue: 2
Page: 363-383
Publish at: 2026-08-01

Production scheduling using a hybrid approach based on genetic algorithm and convergent random search

10.11591/ijeecs.v43.i2.pp534-546
Belbachir Djelloul , Kadri Boufeldja
The flexible job shop scheduling problem (FJSSP) is a highly complex combinatorial optimization problem widely encountered in modern manufacturing systems. Its complexity arises from the simultaneous determination of operation sequencing and machine assignment, making it significantly more challenging than the classical job shop scheduling problem (JSSP). Recent advances in hybrid metaheuristics and intelligent optimization methods have improved solution quality; however, achieving an effective balance between global exploration and local exploitation remains a critical challenge. In this paper, a novel hybrid metaheuristic approach combining a genetic algorithm (GA) and convergent random search (CRS) is proposed to address the FJSSP. The proposed method exploits the global search capability of GA to explore the solution space, while CRS is employed as an adaptive local refinement mechanism applied to elite individuals. This hybridization strategy enhances convergence speed and avoids premature stagnation. Extensive computational experiments are conducted on well-known benchmark instances, including Brandimarte and Kacem datasets. The results indicate that the proposed GA–CRS approach significantly improves the makespan compared to classical GA and PSO based methods. In addition, the algorithm exhibits faster convergence behavior, reaching high-quality solutions in fewer iterations. Statistical analysis using non-parametric tests confirms the superiority of the proposed method. These findings demonstrate that the proposed hybrid GA–CRS algorithm provides a robust and efficient optimization framework for solving large-scale and complex FJSSP instances, outperforming several state-of the-art approaches.
Volume: 43
Issue: 2
Page: 534-546
Publish at: 2026-08-01

Intelligent land use and land cover classification using Sentinel-2 multispectral imagery

10.11591/ijeecs.v43.i2.pp586-594
Neha Vyas , Koushik Sundar , Narayan Vyas
Accurate land use and land cover (LULC) classification is essential for environmental monitoring, agricultural planning, and sustainable resource management. This study investigates the effectiveness of Sentinel-2 multispectral satellite imagery for LULC classification by comparing the performance of three supervised classification algorithms: maximum likelihood classifier (MLC), minimum distance classifier (MDC), and neural networks (NN). Before classification, Sentinel-2 imagery underwent comprehensive preprocessing, including atmospheric correction, radiometric calibration, and cloud masking using ERDAS software to improve image quality and classification reliability. The Villupuram district of Tamil Nadu, India, was selected as the study area due to its diverse land cover characteristics. Classification performance was evaluated using overall accuracy (OA), producer’s accuracy (PA), user’s accuracy (UA), and the Kappa coefficient. Experimental results demonstrate that the MLC achieved the highest OA of 94.81% with a Kappa coefficient of 0.9308, outperforming MDC (90.65%, 0.8753) and NN (84.38%, 0.7917). These findings confirm that supervised classification of Sentinel-2 multispectral imagery provides reliable and accurate LULC mapping, offering valuable geospatial information to support precision agriculture, environmental monitoring, land resource management, and sustainable regional planning.
Volume: 43
Issue: 2
Page: 586-594
Publish at: 2026-08-01

Geospatial data processing and random forest-based intelligent system for regional investment readiness prediction

10.11591/ijeecs.v43.i2.pp651-661
Yudhinanto Cahyo Nugroho , Desmon Desmon , Hasbullah Hasbullah , Triyugo Winarko
This paper presents an intelligent system that integrates geospatial data processing and a random forest (RF) classification model to categorize regional investment readiness (IR). Regional investment planning is often constrained by fragmented socio-economic data, unequal infrastructure distribution, and unquantified disaster risk, which reduce the accuracy of decision making. To address this problem, multidimensional data consisting of socio-economic indicators, infrastructure accessibility, and disaster risk factors were collected from a sample of 15 administrative regions in Lampung Province, Indonesia, and processed through data cleaning, normalization, and feature selection. An IR score was first computed for each region using a weighted composite formula, then discretized into readiness classes and used as the target label to train a RF classifier capable of modeling complex nonlinear relationships among the input features. Given the limited sample size, model performance was evaluated using leave-one-out cross-validation, and classification metrics—accuracy, precision, recall, and F1-score—were reported to assess predictive reliability. The results reveal spatial disparities in IR, where regions with higher human development and better infrastructure tend to exhibit greater investment potential, while areas exposed to higher disaster risk tend to show lower readiness levels. The prediction outputs are integrated into a web-based interactive dashboard that enables spatial visualization and exploration of IR patterns. Given the small and single province sample, the proposed system should be regarded as a preliminary decision-support tool for policymakers and investors to help identify priority regions, and further validation on larger, more geographically diverse datasets is recommended to strengthen generalizability.
Volume: 43
Issue: 2
Page: 651-661
Publish at: 2026-08-01
1 2 3
...
2091 2092
Show 53 of 2092

Discover Our Library

Embark on a journey through our expansive collection of articles and let curiosity lead your path to innovation.

Explore Now
Library 3D Ilustration