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

31,456 Article Results

SqueezeNet-driven AI framework for OCR of handwritten English characters and digits

10.11591/ijeecs.v43.i3.pp928-939
Areen Arabiat , Muneera Altayeb
Optical character recognition (OCR) plays a pivotal role in automated data entry, document digitization, and intelligent user interfaces, particularly as digital content continues to expand across diverse applications. Handwritten character recognition remains a challenging task due to variations in writing styles, noise, and structural complexity. This study investigates the use of artificial intelligence (AI) and machine learning (ML) strategies for classifying handwritten English characters and digits, employing the orange3 platform and a dataset of 2,728 images sourced from Kaggle, partitioned into 70% training and 30% testing sets. A SqueezeNet-based feature extraction approach was implemented to generate 1,000 discriminative features, which were subsequently used to train multiple ML classifiers, including support vector machine (SVM), decision tree (DT), and random forest (RF). Experimental evaluation revealed that the RF classifier achieved the highest accuracy of 99.3%, outperforming DT (90.3%) and SVM (89.9%). These findings highlight the effectiveness of lightweight deep learning architectures such as SqueezeNet in enhancing OCR performance, while demonstrating the robustness of ensemble learning methods in achieving near-perfect classification accuracy. The proposed framework underscores the potential of combining efficient feature extraction with ML classifiers to advance OCR systems for handwritten text recognition in real-world applications.
Volume: 43
Issue: 3
Page: 928-939
Publish at: 2026-09-01

An energy-efficient clustering approach for wireless sensor networks: green hybrid-LEACH

10.11591/ijeecs.v43.i3.pp762-771
Maha I. Raheem , Mazin Ismael Raheem
Energy management is a key challenge in wireless sensor networks (WSNs), primarily due to the limited battery resources of sensors, which shorten the network lifespan. Uneven energy consumption and high routing overhead significantly reduce network longevity. In this paper, we present an energy efficient clustering method called green hybrid-low energy adaptive clustering hierarchy (GH-LEACH). The proposed method selects balanced and energy-efficient cluster heads (CHs) during the setup phase by considering residual energy, node density, node centrality, and distance to the base station (BS). Additionally, an adaptive CH rotation mechanism and multi-hop communication among CHs are incorporated to reduce transmission energy and distribute the network load more effectively. Results demonstrate that GH-LEACH minimizes network overhead, extends the time until the first node dies (FND), improves overall network performance, and enhances throughput compared to existing protocols such as conventional LEACH, LEACH-C, TL-LEACH, and LEACH-F. This approach distributes the energy load among sensors more evenly, especially in denser networks.
Volume: 43
Issue: 3
Page: 762-771
Publish at: 2026-09-01

Aluminium nitride piezoacoustic transducer design for CMOS compatible high-performance ultrasound devices

10.11591/ijeecs.v43.i3.pp718-725
J. Manga , V. J. K. Kishor Sonti
Piezoelectric materials are fundamental to ultrasound probes and scanning devices, enabling the bidirectional conversion of mechanical energy into electrical signals for applications in medical imaging, fingerprint recognition, and non-destructive testing. This capability supports accurate visualization of internal structures, thereby enhancing diagnostic precision. Traditionally, lead zirconate titanate (PZT-4) has been the dominant choice for handheld ultrasound probes due to its high piezoelectric coefficients; however, its limitations include low output efficiency, brittleness, high temperature processing requirements, and significant environmental hazards, particularly marine eco-toxicity during disposal. To overcome these drawbacks, this study investigates aluminium nitride (AlN) as a safer, environmentally sustainable, and thermally stable alternative. AlN offers compatibility with complementary metal oxide semiconductor (CMOS) technology, facilitating integration into compact, high-performance ultrasound devices suitable for biomedical and portable applications. Using COMSOL Multiphysics simulations at 200 kHz, the piezoacoustic response of AlN is analyzed to evaluate its acoustic field behavior under specific electrical inputs. Results highlight AlN’s ability to generate uniform acoustic pressure distributions, supporting its potential for photoacoustic imaging and advanced biomedical diagnostics. This work establishes AlN as a promising material for next-generation ultrasound transducers, combining environmental safety with superior functional performance.
Volume: 43
Issue: 3
Page: 718-725
Publish at: 2026-09-01

Network anomaly detection using hybrid deep learning framework

10.11591/ijeecs.v43.i3.pp792-807
Razibul Islam Khan , Md. Imran Hossain , Hridoy Mohonto , Md. Asmaul Haque , Anwar Parvez Rony , Fazle Rabbi
Network anomaly detection has become increasingly important as modern communication networks face sophisticated and evolving cyber threats that cannot be effectively identified using conventional signature-based intrusion detection systems. Deep learning has demonstrated significant potential for detecting complex attack patterns; however, individual architectures often exhibit limitations in capturing diverse characteristics of network traffic. This paper proposes TriFusion-Net, a hybrid deep learning framework that integrates convolutional neural networks (CNN), long short-term memory (LSTM)-based recurrent neural networks (RNN), and feedforward neural networks (FNN) to improve flow-based network anomaly detection. CNN is employed to extract local spatial patterns from network flow features, LSTM models sequential feature dependencies, and the FNN classifier performs nonlinear decision making for binary intrusion classification. The proposed framework is evaluated using the CIC-IDS-2018 benchmark dataset containing large-scale benign and malicious network traffic. A unified preprocessing pipeline, including data cleaning, feature selection, normalization, and class balancing, is adopted to ensure robust model training and evaluation. Experimental results demonstrate that TriFusion-Net achieves an overall accuracy of 95.66%, with 87.88% precision, 94.99% recall, 91.30% F1-score, and 97.75% receiver operating characteristic area under the curve (ROC-AUC), outperforming the individual deep learning models considered in this study. These results indicate that the proposed hybrid framework provides an effective and computationally efficient solution for real-time network anomaly detection in modern cybersecurity environments.
Volume: 43
Issue: 3
Page: 792-807
Publish at: 2026-09-01

Explainable hybrid machine learning for predicting student enrollment decisions in Indonesian private universities

10.11591/ijeecs.v43.i3.pp880-897
Dybio Dompu Hot Asih , Alex Permatanta Karo-Karo , Erbin Sitorus
Student enrollment prediction is an important challenge for private universities because accurate predictions can support admission planning, marketing strategies, and institutional resource allocation. However, conventional prediction models may provide high predictive performance while offering limited insight into the factors underlying students’ enrollment decisions. This study proposes an explainable hybrid machine learning approach to predict student enrollment decisions in Indonesian private universities and to identify the factors that contribute most strongly to the prediction outcomes. The study uses student enrollment data collected from Universitas Wirahusada Medan and Universitas Audi Indonesia. The proposed approach combines feature selection and multiple machine learning algorithms to improve predictive performance, while explainable artificial intelligence (XAI) using Shapley additive explanations (SHAP) is employed to interpret the model predictions. The models are evaluated using accuracy, precision, recall, F1-score, receiver operating characteristic - area under the curve (ROC-AUC), and confusion matrix analysis. The results show that the proposed hybrid approach provides competitive predictive performance compared with individual machine learning models. The SHAP analysis further reveals the relative contribution and direction of the most influential features in determining student enrollment decisions. These findings demonstrate that combining predictive modeling with explainability can provide not only accurate enrollment predictions but also actionable insights for university decision-makers. The study contributes an integrated approach for supporting data-driven student recruitment and enrollment management in Indonesian private higher education institutions.
Volume: 43
Issue: 3
Page: 880-897
Publish at: 2026-09-01

Mitigation of transmission impairments in MMW-RoF systems using digital compensation techniques: A 17.88 GHz bandwidth study at 160 GHz and 90 GHz

10.11591/ijeecs.v43.i3.pp737-748
Sura Mousa Ali
The unprecedented growth of next-generation wireless communication systems has positioned the transmission of millimeter-wave (MMW) signals over fiber-optic links as a critical research area. As 5G and beyond networks continue to demand ultra-high data rates, massive connectivity, and low latency communication, fiber-wireless convergence has emerged as a key enabling technology for realizing scalable, high-capacity, and cost-effective fronthaul infrastructures. This paper addresses these challenges by designing and simulating an radio-over-fiber (RoF) system using VPIphotonics (Ver. 11.1) and Python (Ver. 3.7.6). The proposed compensation framework incorporates digital signal processing (DSP), Algorithms for dispersion compensation, and a phase noise compensator. The system is planned to operate with a bandwidth of 17.88 GHz, a centre frequency of 160 GHz, and a 90 GHz mmWave. Signal transmission is tested over fibre lengths of 50 km and 70 km, evaluating performance metrics such as error vector magnitude (EVM), signal to noise (SNR), symbol error rate (SER), bit error rate (BER), and data rate consistency for different configurations, including 2x2 MIMO. For 2x2 MIMO, EVM values of approximately 3%, 28.3 dBm SNR, 1E-8 SER, and 2E-09 BER over 50 km and over 70 km channel length achieved 4% EVM, 25.89 dBm SNR, 1E-05 SER, and 2E-06 BER at a data rate of 56.656 Gb/s. This compensator ensured robust performance, enhanced signal integrity, and reliable communication over long distances, meeting the developing demands of modern high-speed communication systems.
Volume: 43
Issue: 3
Page: 737-748
Publish at: 2026-09-01

Explainable machine learning for leukaemia diagnosis: enhancing interpretability and trust

10.11591/ijeecs.v43.i3.pp817-823
Hamza Abu Owida , Mohammad R. Hassan , Nidal Turab , Suhaila Abuowaida
One of the major challenges in applying machine learning (ML) to leukaemia diagnosis is the limited interpretability of deep learning models such as artificial neural networks (ANNs) and convolutional neural networks (CNNs). While these models excel at detecting complex patterns in medical data, they often function as “black boxes,” providing little transparency into their decision-making processes. This lack of explainability raises concerns in clinical practice, where understanding why a model predicts a specific leukaemia subtype is critical for guiding treatment decisions, including chemotherapy and bone marrow transplantation. To address these concerns, researchers are increasingly adopting explainable AI (XAI) techniques. Methods such as saliency maps, feature importance rankings, and attention mechanisms allow clinicians to visualize how specific factors such as genetic mutations, biomarkers, or cell morphology contribute to diagnostic outcomes. These interpretability tools not only enhance transparency but also foster trust among healthcare providers and patients. By bridging the gap between computational accuracy and clinical reliability, XAI ensures that ML models can be safely integrated into diagnostic workflows. Ultimately, explainable approaches strengthen confidence in AI-driven leukaemia diagnosis, supporting more informed medical decisions and improving patient care.
Volume: 43
Issue: 3
Page: 817-823
Publish at: 2026-09-01

Low-depth three-parameter quantum convolutional neural network for image classification

10.11591/ijeecs.v43.i3.pp857-870
Jing Wang , Yong Zhang , Min Zhao
Quantum convolutional neural networks (QCNNs) typically consist of quantum encoding, convolution, and pooling layers. However, existing QCNNs for image classification still face two major limitations. Most encoding layers rely on single-parameter angle encoding, which limits their ability to represent complex image features. Many circuit architectures are relatively deep, increasing cir cuit complexity and noise accumulation and thus hindering implementation on noisy intermediate-scale quantum devices. To address these issues, this paper proposes a low-depth three-parameter encoding quantum convolutional neural network (LTP-QCNN) for image classification. The model employs a three parameter angle encoding layer that maps compressed image features onto the rotation angles of single-qubit gates. A first low-depth quantum convolutional layer is employed to extract local features, followed by a quantum pooling layer for quantum-state dimensionality reduction and information fusion. The pooled quantum states are then processed by a second quantum convolutional layer to extract discriminative features. Finally, the measured quantum feature vector is input into a classical classification layer to produce the final prediction. Simulation experiments were conducted on the Fashion-MNIST, MNIST, and KMNIST benchmark datasets. On Fashion-MNIST, LTP-QCNN achieved an accuracy of 99.85% on binary classification, while the accuracies for five-class and six class tasks reached 90.98% and 89.48%, respectively. The results demonstrate that LTP-QCNN achieves effective and stable image classification under limited quantum resources.
Volume: 43
Issue: 3
Page: 857-870
Publish at: 2026-09-01

Event-level illegal parking detection using YOLOv8 and SSD models

10.11591/ijeecs.v43.i3.pp898-907
Prachi Sodegaonkar , Rahat A. Khan
Unauthorized parking in urban areas results in traffic congestion and leads to the inefficient use of road infrastructure. This study presents an event-level illegal parking detection framework that combines deep learning–based object detection with region-based temporal logic to distinguish transient stops from actual parking violations. A comparative analysis of YOLOv8s and single shot multibox detector (SSD) is conducted using real-world surveillance footage spanning high traffic density and poor lighting conditions. Unlike traditional approaches that rely on hardware-intensive multi-object tracking, our method utilizes a 30-second stationarity threshold within designated regions of interest (ROIs) to detect violations. Experimental results indicate that YOLOv8s outperforms SSD in enforcement reliability, achieving a Recall of 1.00 on a limited-scale evaluation set of 12 confirmed violation events and an F1-score of 0.89, compared with SSD’s Recall of 0.67 and F1-score of 0.77. Moreover, YOLOv8s achieves an inference speed of up to 97 frames per second (FPS), well beyond real-time requirements. These results indicate strong potential for real-time illegal parking detection under urban surveillance conditions.
Volume: 43
Issue: 3
Page: 898-907
Publish at: 2026-09-01

Hybrid supervised–unsupervised machine learning model for proactive congestion control in 5G networks

10.11591/ijeecs.v43.i3.pp772-779
Arun Kodirekka , I. M. V. Krishna , N. Srinivas , Ch. Satya Keerthi N. V. L. , B. Senthilkumaran , Janjhyam Venkata Naga Ramesh
The rapid expansion of fifth-generation (5G) networks has introduced unprecedented challenges in congestion management due to massive device connectivity, diverse traffic patterns, and stringent quality of service (QoS) requirements. Traditional congestion control mechanisms, such as transmission control protocol (TCP), are reactive and often inadequate in dynamic 5G environments, leading to packet loss, latency, and degraded service reliability. To address these limitations, this paper proposes a hybrid congestion control prediction model that integrates supervised and unsupervised machine learning techniques. Supervised learning leverages historical traffic data to predict congestion in known scenarios, while unsupervised learning identifies hidden traffic patterns and anomalies in previously unseen conditions. This dual approach enables proactive congestion detection and adaptive traffic management. Experimental evaluation demonstrates that the hybrid model achieves 95.8% prediction accuracy, reduces detection latency to 180 ms, and lowers false positive rates to 4.5%, outperforming individual supervised or unsupervised models. The results confirm that combining predictive accuracy with anomaly detection provides a robust, scalable, and adaptive solution for congestion control in 5G networks. The proposed model enhances real-time traffic management, ensuring improved QoS, reliability, and resilience in complex 5G environments.
Volume: 43
Issue: 3
Page: 772-779
Publish at: 2026-09-01

Engineering autonomous and resilient power grids for a fully electrified future

10.11591/ijeecs.v43.i3.pp683-694
Tole Sutikno
The global transition toward deep electrification is fundamentally reshaping the role of power grids in modern society. Rapid integration of renewable energy resources, distributed energy systems, electric transportation, and intelligent loads is transforming conventional electricity networks into highly dynamic, decentralized, and cyber-physical infrastructures. In this evolving landscape, traditional power grids designed primarily for reliable electricity delivery are no longer sufficient to address increasing operational complexity, climate-related disruptions, cybersecurity threats, and the growing demand for system flexibility. This editorial argues that the future of electrical power systems depends on the convergence of autonomy and resilience, enabling power grids to anticipate disturbances, adapt to changing operating conditions, coordinate distributed energy resources, and recover rapidly from unexpected events. It discusses recent advances in artificial intelligence, self healing grid technologies, grid-forming power electronics, digital twins, edge intelligence, and cyber-physical security as key enablers of autonomous grid operation. The editorial further highlights emerging research challenges, including human–AI collaboration, AI-native power systems, multi-energy integration, and future regulatory frameworks. Ultimately, autonomous and resilient power grids are presented as the foundational infrastructure supporting a secure, sustainable, and fully electrified future while accelerating the global transition toward carbon-neutral energy systems.
Volume: 43
Issue: 3
Page: 683-694
Publish at: 2026-09-01

Artificial neural network-enhanced solid-state DC breaker for ultra-fast protection in LVDC microgrids

10.11591/ijeecs.v43.i3.pp709-717
Eswaraiah Giddalur , Askani Jaya Laxmi
Low voltage direct current (LVDC) microgrids are increasingly adopted due to their efficiency in integrating distributed energy resources (DERs) and DC loads without multiple conversion stages. However, the presence of high amplitude fault currents and the vulnerability of electronic devices pose significant challenges to reliable protection. Conventional electromechanical switches suffer from slow response times, making them unsuitable for modern LVDC systems. Solid-state circuit breakers (SSCBs), with their millisecond-level interruption capability, offer a promising alternative. This paper introduces an artificial neural network (ANN)-based protection strategy integrated with a bi-directional SSCB for DC bus fault mitigation. The ANN is designed to optimize fault detection and clearing time, achieving rapid isolation within 0.25 ms. A feedforward two-layer neural network with 20 hidden neurons and a sigmoid activation function is trained using multiple algorithms, including Levenberg-Marquardt, Bayesian Regularization, and Scaled Conjugate Gradient. Simulation results demonstrate superior performance compared to conventional differential protection, with the LM algorithm providing the most accurate and efficient fault clearing. The proposed ANN-SSCB approach enhances system reliability, equipment safety, and overall resilience of LVDC microgrids.
Volume: 43
Issue: 3
Page: 709-717
Publish at: 2026-09-01

Dual view explainability-aware log preprocessing for robust anomaly detection toward ER-CyRIS

10.11591/ijeecs.v43.i3.pp871-879
Fathoni Mahardika , Ema Utami , Kusrini Kusrini , Ferry Wahyu Wibowo
Machine learning based intrusion detection can achieve strong benchmark performance yet remain fragile under operational telemetry changes. This paper proposes a dual-view, explainability-aware log preprocessing layer for robust anomaly detection toward ER-CyRIS. The novelty is the use of dynamic-token preservation together with feature stability score (FSS), which turns SHapley additive exPlanations (SHAP)-ranking stability into a preprocessing-level evaluation criterion rather than a post-hoc explanation only. The layer preserves structural log patterns and contextual dynamic tokens, and is evaluated through detection performance, noise degradation, and SHAP-ranking stability. A leakage-controlled ablation on HDFS, BGL, CICIDS2018, and UNSW-NB15 shows that the CICIDS2018 baseline reached F1 = 0.9999 but degraded by 68.1% for XGBoost and 93.9% for random forest under small Gaussian noise. Contextual preprocessing reduced degradation to 58.7%, 54.9%, and 53.6% in selected settings. FSS reached 100% for XGBoost on HDFS and CICIDS2018. The results show that preprocessing mitigates, but does not eliminate, operational brittleness.
Volume: 43
Issue: 3
Page: 871-879
Publish at: 2026-09-01

Zero-trust IoT security using homomorphic encryption, blockchain, and machine learning

10.11591/ijeecs.v43.i3.pp965-974
Khader Musbah Titi
The growing deployment of internet of things (IoT) devices across healthcare, manufacturing, and smart infrastructure has introduced serious security vulnerabilities that can no longer be ignored. Traditional perimeter-based security models have proven inadequate for decentralized IoT environments, especially given that approximately 67% of these devices lack sufficient protection. To address this critical gap, this paper presents a unified security framework that brings together zero trust architecture (ZTA), homomorphic encryption (HE), blockchain, and machine learning (ML) into a single adaptive architecture. At its core, the framework employs Cheon-Kim-Kim-Song (CKKS) based HE for privacy-preserving computation, hyperledger fabric for decentralized identity and key management, and long short-term memory (LSTM) networks for real-time anomaly detection—all operating directly on encrypted data. We validated our approach through extensive experiments on 150,000 samples across 50 IoT devices, and the results were impressive. The framework achieved 97.2% threat detection accuracy with an average response latency of just 120 ms, along with 97.2% resistance to penetration attacks. Notably, it also reduced energy consumption by 25% compared to conventional LSTM implementations. What makes these findings particularly significant is that our framework outperformed both ZTA-only and blockchain-only baselines—by 5.7% and 8.9%, respectively—while maintaining complete data privacy. Furthermore, scalability analyses confirmed linear scaling to over 1,000 devices across 20 edge nodes. Taken together, these results position this framework as a mature, production-ready solution for large-scale IoT security deployments.
Volume: 43
Issue: 3
Page: 965-974
Publish at: 2026-09-01

Seamless UAV integration: a framework for network slicing as a service in 5G and beyond

10.11591/ijeecs.v43.i3.pp749-761
Bouzid Tarek , Noureddine Chaib , Mohamed Lahcen Bensaad
Unmanned aerial vehicles (UAVs) have demonstrated remarkable versatility across monitoring, delivery, and data collection applications. However, the growing complexity of network architectures with 5G and Beyond 5G (B5G) necessitates innovative solutions like network slicing, which faces challenges in radio access network (RAN) efficiency and optimization under higher frequencies. This paper introduces a novel framework for seamless UAV-assisted net work slicing as a service. By integrating UAVs directly into the network infrastructure, the framework enables on-demand delivery of tailored network slices to end users, addressing the limitations of traditional terrestrial approaches while enhancing efficiency and scalability. The proposed framework allows service providers and network administrators to deliver slices more effectively, while specifically tackling high-frequency RAN challenges by deploying UAVs as close-proximity relays to users. A key contribution is complete compatibility and interoperability with existing cellular infrastructure, ensuring seamless integration. The paper presents critical foundational concepts and a literature review, followed by the proposed framework’s architecture, layers, and methodology. Comprehensive tests and comparisons evaluate the framework’s performance, concluding with a discussion of applications and potential enhancements.
Volume: 43
Issue: 3
Page: 749-761
Publish at: 2026-09-01
1 2 3
...
2097 2098
Show 31 of 2098

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