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

Stacking ensemble for multilingual aspect-based sentiment analysis of tourism reviews

10.11591/ijai.v15.i5.pp4381-4391
Basworo Ardi Pramono , Rahmat Gernowo , Aghus Sofwan
Multilingual tourism reviews often express opinions about multiple destination attributes within a single text, making aspect-level sentiment classification challenging. This study compares classical machine learning and ensemble methods for multilabel aspect-based sentiment analysis of amenities, attractions, and accessibility. The dataset comprises 13,085 Indonesian- and English-language Google Reviews labeled through a rule-based procedure. Reviews were represented using term frequency-inverse document frequency features and classified using logistic regression, multinomial naive Bayes, random forest, soft voting, adaptive boosting (AdaBoost), and stacking under a one-vs-rest framework. The stacking classifier combined logistic regression and random forest as base learners with logistic regression as the meta-learner. It achieved the best performance among the evaluated models, with a micro-F1 of 0.8626, macro-F1 of 0.7285, Hamming loss of 0.0633, and subset accuracy of 0.6825. These findings indicate that stacking effectively integrates complementary decision patterns and provides a reproducible framework for multilingual multilabel analysis of tourism reviews.
Volume: 15
Issue: 5
Page: 4381-4391
Publish at: 2026-10-01

Deep multi-block convolutional neural network for quantitative near-infrared spectral analysis

10.11591/ijai.v15.i5.pp4512-4522
Hilmy Abidzar Tawakal , Yohannes Aris Purwanto , Sony Hartono Wijaya , Shelvie Nidya Neyman
Near-infrared (NIR) spectroscopy is a powerful analytical technique, but its effectiveness is often hindered by instrumental noise and a dependency on expert-driven preprocessing and calibration. This study introduces a novel, intelligent deep learning architecture, a deep multi-block convolutional neural network (CNN) that addresses these challenges by performing end-to-end quantitative analysis directly on raw spectral data. By segmenting the NIR spectrum into multiple blocks and processing them in parallel, the model mimics multi-scale feature learning, automatically identifying and leveraging localized spectral regions relevant to target properties. This approach eliminates the need for manual preprocessing, a significant step toward creating autonomous analytical systems. This study evaluated the model on four rice quality datasets (moisture, protein, fatty acids, and total fat) and benchmarked it against traditional partial least squares regression (PLSR) and other deep learning models. The proposed multi-block CNN consistently achieved superior performance, demonstrating its ability to handle raw spectral data robustly. For instance, it achieved coefficient of determination (R2) values of 0.996 ±0.003 for total fat and 0.997 ± 0.001 for fatty acids. Statistical tests confirm that the model’s performance on raw data is not significantly different from its performance with optimized preprocessing, validating its potential for real-world artificial intelligence applications in agriculture, food quality monitoring, and biosystems engineering where rapid, automated analysis is critical.
Volume: 15
Issue: 5
Page: 4512-4522
Publish at: 2026-10-01

Satellite imagery and deep learning framework for highway construction with minimal vegetation impact

10.11591/ijai.v15.i5.pp4933-4945
Meenakshi Meenakshi , Rajesh Thalwagal Matha
Constructing roads through vegetated terrain forces a trade-off between minimizing ecological harm and preserving constructional ease. This work presents a unified, end-to-end framework that reconciles both objectives, with three major contributions. First, propose a hybrid perception model that fuses a graph convolutional network (GCN)—capturing scene-level spatial context—with a capsule network (CapsNet) that encodes part-whole relationships for reliable detection of small, clustered vegetation; this configuration attains an average precision (AP) of 0.925, average recall (AR) of 0.948 (intersection over union (IoU) = 0.5), and 98% land-cover classification accuracy on unseen data. Second, we introduce an ecological-sensitivity cost map fusing vegetation normalized difference vegetation index (NDVI) with terrain slope, over which we evaluate Dijkstra, A*, a novel Eco-Aware A* embedding a vegetation-avoidance penalty, and an non-dominated sorting genetic algorithm II (NSGA-II) multi-objective search that jointly minimizes path length and vegetation damage. Third, on a benchmark modeled on the Bengaluru–Kanakapura road (NDVI = 0.7, slope = 0.3), the eco-sensitive route lowers segment NDVI by roughly 69% and eliminates all high-canopy exposure while adding only about 8% to path length versus an optimized non-vegetated route. Together, these components enable automated, ecologically aware road planning directly from imagery.
Volume: 15
Issue: 5
Page: 4933-4945
Publish at: 2026-10-01

Classifying Arabic handwriting using convolutional neural network

10.11591/ijai.v15.i5.pp4015-4025
Anas Quteishat , Asharaf Al Sharah , Ahmed Qtaishat , Tareq Alawneh
This paper presents a classification of Arabic handwritten letters using convolutional neural networks (CNNs). The paper aims to address the unique challenges of recognizing Arabic letters due to their special characteristics such as dots and similarities. Leveraging deep learning, particularly CNNs, the project employs an architecture specifically designed for image classification tasks, which is well-suited for optical character recognition (OCR) of Arabic handwritten letters. The dataset used consists of images collected from the "Arabic handwritten alphabets, words, and paragraphs" (AHAWP) and "Arabic handwritten character database" (AHCDB), which have been carefully processed to ensure uniformity and diversity. The model was built using deep network designer, a tool in MATLAB, and trained over three epochs to prevent overfitting, achieving an accuracy of 89.2%.
Volume: 15
Issue: 5
Page: 4015-4025
Publish at: 2026-10-01

Deep reinforcement learning-based credit risk prediction model for multi-attribute data

10.11591/ijai.v15.i5.pp4624-4634
Al-Khowarizmi Al-Khowarizmi , Ferry Fachrizal , Romi Fadillah Rahmat , Arif Ridho Lubis , Hariyadi Hariyadi
This paper aims to develop a multi-attribute data-based credit risk prediction model using the deep reinforcement learning (DRL) approach. In the context of creditworthiness evaluation, a model is needed that is not only accurate but also adaptive to complex data dynamics and risk patterns. This study uses a dataset of 5,000 entries that have gone through a preprocessing process, with features such as credit score, income, age, loan amount, and employment history. Initial correlation analysis through heatmap visualization shows that credit score has a strong negative relationship with credit risk, while other features show weak to moderate correlations. The evaluation of model performance is shown through the area under the curve (AUC) value on the receiver operating characteristic (ROC) curve of 0.88, which indicates excellent discrimination ability. The final evaluation results show that the DRL model has an accuracy of 91%, precision of 89%, recall of 90%, and F1-score of 89.5%, indicating that the model not only has high accuracy but is also balanced in its sensitivity and specificity. Thus, the application of DRL has proven effective in predicting credit risk adaptively and accurately, and can be implemented as an intelligent solution in an automated credit decision-making system based on multi-attribute data.
Volume: 15
Issue: 5
Page: 4624-4634
Publish at: 2026-10-01

Interpretable machine learning framework for congestive heart failure risk prediction

10.11591/ijai.v15.i5.pp4345-4355
Sakchai Tangprasert , Nalinpat Bhumpenpein , Siranee Nuchitprasitchai , Yuenyong Nilsiam
Congestive heart failure (CHF) is a leading cause of morbidity and mortality worldwide and is influenced by multiple clinical risk factors, including age, sex, blood pressure, cholesterol levels, and electrocardiographic characteristics. This study employs publicly available heart disease datasets compiled from populations in Cleveland, Hungary, Switzerland, and Long Beach repositories (2015–2020) and analyzes 11 predictor attributes and one binary outcome using the cross-industry standard process for data mining (CRISP-DM) framework. Four supervised machine learning (ML) models—decision tree, random forest, gradient boosted trees, and naïve Bayes—were developed and evaluated using confusion matrix–based performance metrics, with results showing that the decision tree model offers balanced predictive performance and clear interpretability, making it suitable for heart failure risk prediction. The findings demonstrate the potential of ML approaches to support clinical risk assessment and enhance decision-making in heart failure management.
Volume: 15
Issue: 5
Page: 4345-4355
Publish at: 2026-10-01

Edge-aware adaptive partial differential equation image inpainting for high-quality restoration

10.12928/telkomnika.v24i5.27914
Fouzia; University of Mohamed El Bachir El Ibrahimi Hamadache , Zine-Eddine; Mohamed Khider University of Biskra Baarir , Nacira; University of Mohamed El Bachir El Ibrahimi Diffellah , Abdelmalik; IEMN UMR CNRS 8520 Ahmed-Taleb , Tewfik; University of Ferhat Abbas Bekkouche , Rabah; University of Mohamed El Bachir El Ibrahimi Hamdini
Image inpainting aims to reconstruct missing or damaged regions in digital images while preserving structural consistency and visual coherence. Among partial differential equation (PDE)-based approaches, the model introduced by Marcelo Bertalmio is a foundational technique for isophote-driven structure propagation. However, its performance often degrades in large damaged re gions. This limitation is mainly caused by insufficient transport strength and isotropic diffusion, which lead to structural blurring. In this paper, we propose an edge-aware adaptive PDE inpainting model that introduces a transport am plification factor and a spatially-adaptive diffusion term to improve edge prop agation and structural preservation. Experimental results on various test im ages demonstrate that the proposed approach improves structural reconstruction. Quantitative evaluation using peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) metrics confirms higher restoration accuracy and better edge preservation compared with baseline methods.
Volume: 24
Issue: 5
Page: 1710-1720
Publish at: 2026-10-01

Optimization of MSMEs promotion strategy using GIS with random forest algorithm and A-star algorithm

10.12928/telkomnika.v24i5.26692
Sulyono; Darmajaya Institute of Informatics and Business Sulyono , Suci; Darmajaya Institute of Informatics and Business Mutiara , Agus; Darmajaya Institute of Informatics and Business Rahardi , Sri; Darmajaya Institute of Informatics and Business Lestari , Yulmaini; Darmajaya Institute of Informatics and Business Yulmaini , Ruki Rizal; Darmajaya Institute of Informatics and Business Nul Fikri , Ali; Darmajaya Institute of Informatics and Business Syarifuddin
Micro, small, and medium enterprises (MSMEs) make a main contribution to improving a country’s economy, but it still faces various problems including limited market access, ineffective promotion, and lack of understanding in business profiles. The purpose of this study is to design a geographic information system (GIS) combined with the random forest (RF) algorithm and the A-star algorithm. The RF algorithm functions are to classify MSMEs. Meanwhile, the A-star algorithm is to recommend the shortest route to support accessibility. In addition, it is equipped with a dashboard and MSMEs profile as a reference for stakeholders in making MSMEs development policies and as information to the public in order to expand the reach of promotion in introducing MSMEs and attracting investors, thus optimizing MSMEs promotion strategies. In addition, the evaluation results of the classification model show an accuracy value of 99.6%, indicating that the model is able to predict very well, making it highly recommended for MSME classification.
Volume: 24
Issue: 5
Page: 1526-1534
Publish at: 2026-10-01

Virtualized evolved packet core deployment strategies in emerging Palestinian mobile networks

10.12928/telkomnika.v24i5.27849
Saed; An-Najah National University Tarapiah , Shadi; University of Dubai Atalla , Khawla; Ooredoo Palestine Daraghme
Mobile operators in emerging markets face rising data demand under revenue, infrastructure, regulatory, and investment constraints. Network function vir tualization (NFV) enables evolved packet core (EPC) modernization, but op erators must determine which deployment model provides the best strategic and operational fit. This paper examines EPC virtualization as an operator level deployment-selection problem using Palestine as a case study. A context specific decision framework integrates operator readiness, stakeholder-derived key performance indicator (KPI) weights, deployment-model ratings, capital expenditure (CapEx) / operating expenditure (OpEx) analysis, analytic hierar chy process (AHP) consistency checking, and technique for order of preference by similarity to ideal solution (TOPSIS)-style robustness validation. Five al ternatives are compared under the same operator scenario: owned legacy EPC, owned NFV/virtualized evolved packet core (vEPC), network functions virtual ization infrastructure as a service (NFVIaaS), virtual network platform as a ser vice (VNPaaS), and virtual network service as a service (VNSaaS). The novelty lies in linking multi-criteria scoring to local regulatory, infrastructure, security, supply-chain, and business constraints rather than applying generic weighted ranking alone. Scenario-based technical validation examines latency / quality of service (QoS) exposure, scalability, recovery, elasticity, and observability. The results indicate that fully owned vEPC provides the strongest overall fit when service control, quality, implementation constraints, and long-term operational efficiency are jointly considered.
Volume: 24
Issue: 5
Page: 1480-1491
Publish at: 2026-10-01

MRI segmentation and classification using 1-D and 2-D empirical mode decomposition

10.12928/telkomnika.v24i5.27976
Giang Hong; University of Science Nguyen , Yen Thi Hoang; University of Science Hua , Ngan Vuong Thuy; University of Science Nguyen , Liet Dang; Dong Nai University Van
This study proposes a new framework for brain magnetic resonance imaging (MRI) tumor detection and classification using 1-D and 2-D empirical mode decomposition (EMD) / bidimensional empirical mode decomposition (BEMD). These adaptive algorithms optimize image denoising, segmentation and feature extraction. Tumors are detected using a global threshold derived from EMD-generated bimodal curves. A visual geometry group 19 (VGG19) model extracts features from the regions of interest (ROI) across both denoised images and BEMD components. The method was validated on the public Figshare brain tumor dataset. For segmentation, the proposed method performed similarly to multilevel Otsu, Kapur-Harmonic, and K-means algorithms. It achieved an average precision of 96.70%, slightly trailing K-means of 97.07%, and a sensitivity of 91.07%, outperforming K-means but falling behind Kapur and Otsu. Notably, it achieved the highest F1-score of 93.70%, tying the Otsu technique. These results highlight enhanced tumor localization, reflecting segmentation that closely aligns with the ground truth. For tumor classification, ensemble learning was used to develop nine models, all of which surpassed 93% accuracy; one achieved 100%, and three reached 96.7%. The proposed method outperformed seven state-of-the-art techniques with a peak accuracy of 99.02%, demonstrating its clinical viability. Overall, integrating EMD and BEMD algorithms provides a superior framework for brain tumor segmentation and classification compared traditional methods.
Volume: 24
Issue: 5
Page: 1728-1744
Publish at: 2026-10-01

Scaling of sustainable development goals in science education using Thurstone’s pairwise comparison–based scaling technique

10.11591/ijere.v15i5.39770
Ahmet Volkan Yüzüak , Zekiye Merve Öcal , Emrah Hiğde
A lack of evidence-based guidelines on which sustainable development goals (SDGs) need to be focused on in educational settings can be regarded as the key deficiency that leads to an inefficient sustainability learning process. As sustainability-related issues become ever more acute, there is a necessity to determine which SDGs should be prioritized pedagogically. Thus, the current study aims at examining how future teachers evaluate the significance of selected SDGs in relation to environmental learning. Applying the paired comparison methodology, the authors gathered data from 129 third- and fourth-year prospective teachers concerning their perceptions of 11 SDGs (SDG 1, 2, 3, 4, 5, 6, 7, 12, 13, 14, and 15). According to the results, SDG 3, SDG 6, and SDG 2 seem to be the most important goals, while SDG 15, SDG 5, and SDG 14 appear less critical. Thus, one may hypothesize about hierarchy within SDGs based on their importance as prerequisites for environmental problems solving. In terms of practical application, this paper offers a definite methodology that instructors can apply to rank and arrange SDGs for teaching purposes, as well as useful tips for curriculum designers and policymakers who seek to develop sustainability education programs.
Volume: 15
Issue: 5
Page: 3716-3725
Publish at: 2026-10-01

Comparative deep learning architectures for railway train detection using accelerometer

10.12928/telkomnika.v24i5.27752
Heri; Universitas Muhammadiyah Lamongan Ardiansyah , Denny; Universitas Muhammadiyah Gresik Irawan , Dhio; Universitas Muhammadiyah Lamongan Ramadhon Saripudin , Gilang; Universitas Muhammadiyah Lamongan Septia Firmansyah , Rohmatin; Universitas Muhammadiyah Gresik Agustina
Railway level crossing accidents remain a critical safety concern worldwide, necessitating reliable train detection systems particularly at unguarded crossings. This study presents comparative evaluations from four deep learning (DL) architectures convolutional neural network (2D-CNN), long short-term memory (LSTM), MobileNetV2, and hybrid CNN-LSTM of railway train detection using triaxial accelerometer sensor data. Vibration signals acquired from piezoelectric sensors mounted on railway infrastructure were transformed into spectrogram representations and classified into four proximity categories: idle, far-distance (>400 m), near distance (200-400 m), and train passing (
Volume: 24
Issue: 5
Page: 1755-1767
Publish at: 2026-10-01

Beamforming design for interference cancellation in fifth generation new radio system

10.12928/telkomnika.v24i5.27664
Heba; Damascus University Shafeik Kashour , Abdelrazak; Damascus University Badawieh , Ali; Tartous University Ahmad , Mohammad; Higher Institute for Applied Science and Technology (HIAST) Rizk Assaf
This study proposes a hybrid beamforming (HBF) architecture for interference mitigation design aimed at improving interference cancellation in fifth-generation new radio (5G-NR) systems. The proposed architecture employs interference suppression at two processing level: the linear eigen zero forcing (EZF) method is applied at the base station (BS) to mitigate inter-user interference (IUI), while the nonlinear successive interference cancellation (SIC) is implemented at each user terminal to cancel intra-user interference. Although EZF is linear, meaning it doesn’t require iterative or cumulative operations, its operations are complex and rely on matrix inversion. On the other hand, SIC is a non-linear algorithm, meaning it requires many sequential iterative operations, which increases the probability of cumulative errors. Therefore, using EZF as the first stage makes the signals to be processed in SIC less overlapping, thus reducing cumulative errors, especially under varying channel conditions. The single to interference and noise ratio (SINR) reached a value of 11.9 dB. This is based on the simulation scenarios that were tested. and a channel quality indicator (CQI) value of 10. Moreover, the proposed scheme provides a spectral efficiency of 36 bit/s/Hz at 100 MHz bandwidth and -20 dB signal-to-noise ratio (SNR), achieving performance approaches the performance of a fully digital interference-free system under the considered simulation conditions. while maintaining practical computational complexity.
Volume: 24
Issue: 5
Page: 1469-1479
Publish at: 2026-10-01

Solid oxide fuel cells as next‑generation high‑efficiency alternatives to combined cycle gas turbines

10.12928/telkomnika.v24i5.27887
Ignatius; Politeknik Negeri Bandung Riyadi Mardiyanto , Handrea; Politeknik Negeri Bandung Bernando Tambunan , Haryadi; Politeknik Negeri Bandung Haryadi , Lina; Politeknik Negeri Bandung Troskialina , Sri; Politeknik Negeri Bandung Paryanto Mursid , Yusuf; Politeknik Negeri Bandung Sofyan , Lidya; Politeknik Negeri Bandung Elizabeth
This study provides a direct thermodynamic comparison between a combined cycle gas turbine (CCGT) and a solid oxide fuel cell (SOFC) system, each designed for approximately 450 MW gross output. The benchmark CCGT, based on Siemens STG5‑4000F and SST5‑5000 units, achieved 61% gross efficiency under baseline conditions, with moderate sensitivity to fuel composition. In contrast, the SOFC system, modeled with an active cell area of 250,000 m², demonstrated superior performance, reaching nearly 70% gross efficiency and exceeding 63% net efficiency, while requiring lower fuel input (18.5 kg/s versus 19.5 kg/s for CCGT). Efficiency trends highlight fundamental differences: CCGT shows incremental gains with increased fuel flow but remains below 57% net efficiency, whereas SOFC sustains higher values despite greater sensitivity to input variations. These findings emphasize SOFC’s potential as a next‑generation utility‑scale power technology, offering higher efficiency and reduced fuel demand, while acknowledging challenges in scalability, durability, and cost that must be addressed for practical deployment.
Volume: 24
Issue: 5
Page: 1805-1815
Publish at: 2026-10-01

High-performance 28 GHz antenna array design for millimeter wave 5G systems

10.12928/telkomnika.v24i5.27857
Moussab; University Abdelmalek Essaâdi Chbeine , Abderrahman; University Abdelmalek Essaâdi Chbeine , Mohamed; University Abdelmalek Essaâdi Bayjja , Lahcen; Hassan II University of Casablanca Sellak
This paper presents the design and performance evaluation of compact 28 GHz microstrip antenna arrays for millimeter-wave (mmWave) fifth generation (5G) wireless communication applications. The proposed antennas are implemented on a Rogers RT/duroid 5880 substrate with a relative permittivity of 2.2 and a thickness of 0.508 mm and were evaluated using full-wave electromagnetic simulations in Computer Simulation Technology (CST) Microwave Studio. Starting from a single antenna element, several scalable array configurations, including 1×2, 1×4, 1×8, and 2×8 structures, were designed to improve antenna gain and radiation directivity while maintaining good impedance matching characteristics. The novelty of this work lies in the development of compact scalable antenna arrays using a simple feeding network configuration to achieve progressive gain enhancement with reduced structural complexity. The simulation results show that the antenna performance improves as the number of array elements increases, achieving a maximum gain of 16.2 dB at 28 GHz with satisfactory return loss, voltage standing wave ratio (VSWR), and radiation behavior. In addition, the proposed arrays provide an appropriate trade-off between gain, bandwidth, and compact size, making them suitable candidates for future high-data-rate mmWave 5G communication systems.
Volume: 24
Issue: 5
Page: 1492-1503
Publish at: 2026-10-01
1 2 3
...
2097 2098
Show 18 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