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

Fine-tuning language models for social sentiment: insights from startup user feedback

10.11591/ijai.v15.i5.pp4432-4446
Li-Hua Li , Agus Cahyo Nugroho
Startup growth and sustainability are heavily impacted by user comments. This research has done a comparative study between large language models (LLMs), small language models (SLMs), and traditional machine learning (ML) methods in classifying user comments on startups. The performances of bidirectional encoder representations from transformers (BERT), Indonesian bidirectional encoder representations from transformers (IndoBERT), distilled bidirectional encoder representations from transformers (DistilBERT), and ML models; logistic regression, support vector machine (SVM), random forest, and extreme gradient boosting (XGBoost) were evaluated with a diverse dataset and beyond extensive preprocessing. Out of all the models this research trained, IndoBERT outperformed with an accuracy and recall 90.99% and F1-score of 89.94%, due to its intermediate pretraining for Indonesian language task. DistilBERT, from the SLM family of models and only requiring a fraction of resources to make predictions performed significantly better (over 90% in both accuracy and recall). These outcomes indicate that LLMs, particularly those targeting specific languages, are significantly superior in performance for sentiment analysis. The results seem to imply that models like these can be utilized to study the phenomenon and further enhance user engagement with startups. Future work will involve the further development of more language specific BERT versions and expanding LLMs in other ways within startup spaces.
Volume: 15
Issue: 5
Page: 4432-4446
Publish at: 2026-10-01

Traffic flow optimization using reinforcement and thread-based V2X communication for smart traffic management

10.11591/ijai.v15.i5.pp4061-4070
Parul Arora , Ritika Wason , Devansh Arora , M. N. Hoda , Sumit Gupta
This research proposes an advanced framework that integrates reinforcement learning techniques with thread-based vehicle-to-everything (V2X) communication protocol to address urban traffic jams, which lead to increased delay, inefficient fuel consumption, and deteriorating air quality, posing a significant challenge to sustainable living. This framework empowers cars to “talk” directly to each other as well as with traffic signals, adapting to the real-time flow on the roads. By placing the decision-making process closer to where it’s needed, right at the network’s edge, the system is able to respond quickly to changes in traffic without relying on the usual pre-set signal cycles. Unlike many traditional traffic management solutions that are slow to adjust, the proposed framework allows vehicles and intersections to coordinate in real time. This reduces unnecessary waiting, smooths out traffic flow, and helps cut down on emissions. Through detailed simulations using the simulation of urban mobility (SUMO) traffic model, it was found that this system led to significant improvements: average waiting time was reduced by over 40%, queue lengths were cut by more than half, and vehicle throughput increased by nearly 30%. Additionally, fuel consumption dropped by approximately 24%, and carbon dioxide (CO2) emissions were reduced by more than 25% compared to existing systems.
Volume: 15
Issue: 5
Page: 4061-4070
Publish at: 2026-10-01

Robust trading in stochastic chaos: a hybrid fuzzy multi-expert approach

10.11591/ijai.v15.i5.pp4105-4112
Andrey Musayev , Dmitry Grigoriev
Classical trend-following strategies often exhibit unstable performance in the EUR/USD foreign exchange market due to filter lag, noise-induced false signals, and abrupt regime transitions associated with chaotic price dynamics. This study proposes a hybrid fuzzy multi-expert framework that integrates bidirectional exponential filtering for low-latency trend extraction with a Mamdani-type fuzzy inference system that jointly evaluates quantitative trend indicators and qualitative information derived from economic news and textual forecasts. The proposed approach is evaluated on minute-level EUR/USD data over 10-, 25-, and 100-day test windows. Empirical results show that the fuzzy consensus mechanism improves robustness under non-stationary conditions. In a representative 25-day evaluation period, net profit increases from 182 to 560 pips, the win rate improves from 55% to 67%, and the maximum drawdown decreases from 118 to 38 pips. These findings suggest that incorporating linguistic uncertainty and sentiment-like contextual cues into technical trading signals can mitigate spurious trade activation and stabilize decision-making in stochastic and chaotic market environments, offering a practical route toward more resilient algorithmic trading systems.
Volume: 15
Issue: 5
Page: 4105-4112
Publish at: 2026-10-01

Exploring barriers to gamification implementation in classrooms: an empirical investigation of pre-service teachers’ experiences

10.11591/ijere.v15i5.39061
Yaggadi Sai Kumar , Ashalatha Shanthipalla , Suresh Anuganti , Chintapalli Sowmya Nakshatram , Gajula Vinodkumar
Gamification has emerged as an effective teaching approach to enhance student engagement and achievement outcomes in the learning process. However, it is implementation remains limited due to challenges faced in classrooms, particularly among pre-service teachers, who face unique, understudied barriers in their teaching internships, compounded by insufficient institutional support and inadequate practical preparation. This study employed a quantitative descriptive survey design to investigate the curriculum, resource, and training barriers perceived by pre-service teachers during their internships. A total of 347 pre-service teachers enrolled in the Bachelor of Education (B.Ed.) program participated in the study, after their 16-week teaching internships in secondary schools. Data were collected using a 25-item Likert scale across three barrier dimensions. Descriptive statistics revealed that insufficient mentor guidance constituted the most critical training barrier (M=3.919, SD=1.031), time constraints dominated resource barriers (M=3.876, SD=1.104), and mentor adherence to traditional approaches represented the foremost curriculum barrier (M=3.885, SD=1.064). The results underscore the need for comprehensive reforms in curriculum policy, strengthening technological infrastructure, and redesigning teacher education programs with practical gamification training. Addressing these systemic barriers is essential for preparing future teachers to leverage gamification as a viable teaching methodology in contemporary classrooms.
Volume: 15
Issue: 5
Page: 4014-4025
Publish at: 2026-10-01

The impact of digital comics as a teaching and learning tool in economics education on future-ready talent development: the mediating role of attitude

10.11591/ijere.v15i5.37622
Hainnuraqma Rahim , Abd Hadi Mustaffa , Norashida Othman , Putri Aliah Mohd Hidzir
Economics education at the tertiary level dominated by text-heavy materials that often disengage students. At the same time, the potential of digital comics as an innovative pedagogical alternative are insufficiently examine. Hence, this study examines the impact of digital comics (IMP) on students’ learning in economic subjects at Universiti Teknologi MARA (UiTM) Melaka, with particular focus on the mediating role of attitude (ATT). A quantitative design was employed using structured questionnaires distributed to 146 diploma and degree students to measure the determinants of perceived IMP. Data were analyzed using the partial least squares– structural equation modeling (PLS-SEM). The findings show that ATT and subjective norms (SN) significantly influence the IMP, while digital literacy (DL) and perceived behavioral control (PBC) exhibit no direct effects. In addition, ATT significantly mediates the relationships with SN, PBC and the IMP. Practically, the findings suggest that lecturers can embed digital comics into case-based class activities to enhance engagement and conceptual understanding. The novelty of the study indicated that this is the first studies applying theory of planned behavior (TPB) to digital comics in economic education. This study also offers future research which can be extend the studies into experimental, longitudinal and cross-sectional.
Volume: 15
Issue: 5
Page: 4192-4206
Publish at: 2026-10-01

Development model of remote sensing data management for long-term storage using big data

10.12928/telkomnika.v24i5.27639
Riyan; National Research and Innovation Agency Mahendra Saputra , Agnes; National Research and Innovation Agency Sondita Payani , Ogi; National Research and Innovation Agency Gumelar
The growing volume of remote sensing data requires efficient approaches for long-term storage and data management. The National Remote Sensing Data Bank (BDPJN) stores multi-resolution satellite imagery acquired since 1993 to support disaster monitoring, climate studies, natural resource management, and other Earth observation applications. However, managing continuously growing archives remains challenging due to manual migration processes, inefficient storage utilization, and limited scalability. This study proposes an integrated remote sensing data management model that combines hierarchical storage management (HSM) with Apache Spark based parallel processing to support policy-driven data migration and long term storage optimization. The proposed model enables data migration across multiple storage tiers based on data characteristics and storage policies while improving migration efficiency through distributed processing. Experimental evaluation was conducted using Landsat-8, SPOT 6, SPOT-7, and Pleiades datasets under both 1 Gbps and 10 Gbps network environments. The results demonstrate migration performance improvements ranging from 80.14% to 91.04% compared with the existing manual migration system. These findings indicate that integrating HSM with Spark-based parallel processing can reduce migration bottlenecks, improve storage utilization, and provide a scalable and cost-effective approach for managing large-scale remote sensing archives while supporting long-term data governance and accessibility.
Volume: 24
Issue: 5
Page: 1535-1549
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

Review on badminton players hand-grip analysis and tactical dynamics prediction using egocentric vision

10.12928/telkomnika.v24i5.27915
Sriramachandra; Alliance School of Advanced Computing Murthy K. , Shekhar; Alliance School of Advanced Computing R. , Athira B.; Alliance School of Advanced Computing Kaimal
Many works have already been carried out on computer vision-based sports analytics, activities in egocentric vision to improve player performance. But less is explored in terms of vision based kinematic analysis in badminton where shuttle trajectory is explored, angles are derived from features and analyzing biomechanics of the player first person vision cameras. This paper reviews existing research on badminton hand-grip analysis and tactical prediction using egocentric vision. It highlights key methods, datasets, and challenges in applying computer vision techniques to sports analytics. The review focuses on computer vision approaches for analyzing hand-grip patterns and predicting tactical dynamics in badminton using egocentric video data. The study summarizes current advancements, identifies research gaps, and outlines future directions for improving egocentric-vision-based sports analytics.
Volume: 24
Issue: 5
Page: 1721-1727
Publish at: 2026-10-01

Exploration of thesis topic trends of students majoring in informatics and computer engineering with LDA method

10.12928/telkomnika.v24i5.27674
Ruslan; University of Makassar Ruslan , Rezki Nurul; University of Makassar Jariah S. Intam , Sasmita; University of Makassar Sasmita , Dewi; University of Makassar Fatmarani Surianto , Andi Akram; University of Makassar Nur Risal , Nur Azizah; University of Makassar Eka Budiarti
The selection of a thesis topic is very important because it determines the focus of the research and its contribution to knowledge. However, many students find it difficult to choose a topic that suits their interests and expertise. This study models the thesis topics of students in the Department of Informatics and Computer Engineering (JTIK) using the latent Dirichlet allocation (LDA) method, with term frequency–inverse document frequency (TF–IDF) as the model input. The data set includes 969 thesis titles from 2009 to 2024. The optimized LDA model identifies 17 main topics by adjusting parameters such as the number of topics, alpha, and beta. The best coherence value (0.7431) is achieved with alpha = 0.81, beta = 0.01, and 17 topics. The dominant themes included information system development, computer networks, and multimedia, reflecting the main research areas of JTIK. In addition, a web-based system was developed and integrated with the best model to help students identify relevant topics and find thesis references. This study demonstrates the effectiveness of topic modeling in higher education and provides insights into academic research trends.
Volume: 24
Issue: 5
Page: 1550-1560
Publish at: 2026-10-01

Explainable AI-based hybrid GNN-MLP model for strawberry fruit disease detection using hyperspectral imaging

10.11591/ijict.v15i3.pp1364-1375
Varsha Kiran Bhosale , Chin-Shiuh Shieh
Strawberries are severely affected by the main fungal diseases such as anthracnose fruit rot, grey mould, and powdery mildew, directly reducing commercial yield and post-harvest quality. In this paper, we propose an enhanced hybrid deep-learning method by combining graph neural networks (GNNs) and multi-layer perceptrons (MLPs) for effective strawberry disease detection in real environments of fields. This dataset contains images acquired from both public domain repositories as well as farm operational settings, encompassing three classes of diseases with varying lighting conditions, background noise, and occlusion. A structured pre-processing workflow comprising contrast enhancement, denoising, and synthetic hyperspectral simulation is used to enhance subtle lesion features and stabilize the subsequent feature extraction. The hybrid GNN–MLP framework possesses the merits of spatially local lesion topology and two-point contextual information, which can improve disease classification compared with common CNN-based structures. 5-fold cross-validation shows that the model holds 93.59% accuracy and a macro F1-score of 90.39%, showing good generalization even with heterogeneous input regimes. Transparent models use local interpretable model-agnostic explanations (LIME) and gradient-weighted class activation mapping (Grad-CAM) in combination, highlighting disease-relevant areas that give an interpretable rationale for each prediction. To conclude, the developed system offers an accurate and explainable solution that has a computationally efficient commitment for real-time monitoring of disease in smart agriculture settings, particularly on low-cost hardware assets.
Volume: 15
Issue: 3
Page: 1364-1375
Publish at: 2026-09-01

Classroom behavior mining in adolescents: a cognitive and data-driven approach using BEHAVE_Apriori

10.11591/ijict.v15i3.pp1058-1065
Suresh Govindarajalu , Muthukumaran Subramaniyan , Kamatchy Balakrishnan , Kalaichelvi Nagarajan , Nandhini Krishnamoorthy
Adolescence is a critical developmental stage that leads to essential changes in social, emotional, and cognitive domains that affect conduct in the classroom. Students' perceptions, processing, and reactions to their learning environment are better-understood thanks to cognitive psychology. However, contemporary data mining techniques frequently ignore the environmental, emotional, and cognitive elements influencing teenage behavior in learning environments. This research presents a comprehensive approach to analyzing teenage college students' classroom behavior by integrating cognitive psychology with data-driven methods to identify key behavioral traits shaped by both external and internal factors. A brand-new algorithm called the behavioral evaluation via hybrid attributes and valuable extraction using the Apriori (BEHAVE_Apriori) approach is presented. Also, a variety of feature selection (FS) strategies, including information gain (IG), chi-squared (CS), and tree-based approaches, are used for FS. Then, using the Apriori algorithm, association rules are found that relate behavior patterns to elements like family history, academic involvement, and peer influence. The IG-based FS combined with the Apriori algorithm delivered the best performance, generating 95 rules in 0.0241 seconds, outperforming CS (154 rules, 0.0629s) and tree-based FS (251 rules, 0.1394s), while the unfiltered dataset produced 514 rules in 0.2853 seconds.
Volume: 15
Issue: 3
Page: 1058-1065
Publish at: 2026-09-01

An effective blockchain-based system for tracking viral content origins in social media

10.11591/ijict.v15i3.pp1313-1321
Nageswararao Sirisala , Srinivasulu Sirisala , Nalavala Ramanjaneya Reddy , Syed Anisha
Tracking the origins of viral content in social media is crucial for identifying misinformation, ensuring accountability, and protecting intellectual property. By tracing content back to its source, platforms can curb the spread of false narratives, hold malicious actors responsible, and safeguard creators’ rights. In this work, “Effective blockchain based system for tracking viral content origins in social media-(BCTVCO)” is proposed. The BCTVCO is a blockchain-based content authentication platform that leverages interplanetary file system (IPFS), smart contracts, and cryptographic hashing to verify digital assets and detect unauthorized reuse. This decentralized application addresses the challenges of content authenticity, ownership verification, and intellectual property protection in the digital space. The system integrates Ethereum smart contracts (Solidity) to store immutable content records and uses SHA-256 hashing for secure file integrity verification. Content is uploaded to IPFS via Web3. Storage, ensuring distributed and tamper-resistant storage. The React-based frontend with MetaMask authentication allows users to seamlessly register, upload, and track their content. In implementation, BCTVCO outperformed existing methods and proved to be a scalable, transparent, and secure blockchain-based content verification system. Future enhancements of BCTVCO involve multi-chain support and AI-powered content analysis to strengthen security and usability.
Volume: 15
Issue: 3
Page: 1313-1321
Publish at: 2026-09-01

Cross-modal attention fusion using vision transformers for robust student attentiveness estimation

10.11591/ijict.v15i3.pp935-943
Rajasekaran Mariswamy , Praveen Sundar
Automated student attentiveness estimation is a fundamental component of intelligent e-learning systems and adaptive classroom analytics. Traditional convolutional and recurrent architectures often struggle to model long-range temporal dependencies and complex inter-modal relationships inherent in engagement behavior. To address these limitations, this paper proposes a cross-modal attention fusion framework built upon a vision transformer (ViT) backbone for robust student attentiveness estimation. The proposed architecture leverages patch-based visual encoding through a ViT to capture global spatial dependencies, while behavioral cues such as gaze direction, head pose, and blink dynamics are embedded into a shared latent representation space. A cross-modal multi-head attention mechanism is introduced to dynamically learn interactions between visual and behavioral modalities, replacing static weighted fusion strategies. Temporal dynamics are modeled using a Transformer encoder, enabling effective long-range sequence modeling without recurrent dependencies. Experimental evaluation on a benchmark attentiveness dataset demonstrates superior performance compared to CNN–LSTM-based models, achieving improved accuracy, F1 score, and robustness under challenging lighting and occlusion conditions. Ablation studies validate the contribution of cross-modal attention and transformer-based temporal modeling. The proposed framework maintains real-time feasibility while significantly enhancing discriminative capability.
Volume: 15
Issue: 3
Page: 935-943
Publish at: 2026-09-01

QSUMeMarket: a decision support system framework for smes to process customer orders

10.11591/ijict.v15i3.pp975-985
Winston G. Domingo , Jennifer A. Gamay , Virdi C. Gonzales , Selino S. Malunao
With the progression of big data analytics (BDA), point of sale (PoS) could be amalgamated with an inventory management (IM) system. The problem of SMEs is to generate a financial report only using simple calculations based on income and expenses. To develop a web-based computerized system to solve the problems encountered and ease the marketing office’s operations. The used of a descriptive research design and software development life cycle (SDLC) methodology. The ISO/EIC 25010:2011 software quality standard was chosen as one of the most comprehensive software-quality evaluation models. The software performs exceptionally well in several key areas, including performance efficiency (4.00), usability (3.94), portability (3.89), compatibility (3.83), security (3.80), maintainability (3.70), and reliability (3.67). These ratings highlight its strengths in speed, security, and overall user experience. However, with the lowest score being 3.44, there is still potential for improvement to better align the software with user needs and business objectives. While it meets industry standards and is suitable for practical use, ongoing enhancements are essential to sustain and elevate its quality.
Volume: 15
Issue: 3
Page: 975-985
Publish at: 2026-09-01

Digital forensics for cultural preservation: multi-device image classification of the historic Surabaya City Hall

10.11591/ijict.v15i3.pp995-1003
Ulfa Meilinda Putri , Imam Yuadi
Cultural heritage preservation increasingly relies on digital forensics to ensure authenticity and consistency in heritage documentation. This study presents a digital forensic framework based on machine learning for classifying multi-device images of the historic Surabaya City Hall. The dataset was collected from nine smartphone devices and preprocessed through standardization, 360° rotational augmentation, and three filtering methods: gaussian, median, and laplacian. Three supervised algorithms (support vector machine (SVM), K-nearest neighbor (KNN), and logistic regression (LR)) were evaluated using accuracy, macro average, and weighted average of precision, recall, and F1-score. The results indicate that image preprocessing substantially affects model performance, with the gaussian-filtered KNN achieving the best result, reaching 92% accuracy, and balanced macro and weighted F1-scores of 0.92-0.93. Confusion-matrix analysis revealed minor misclassifications among iPhone models with similar sensor characteristics, while other devices were accurately identified. The findings confirm that gaussian filtering improves feature consistency and that KNN’s distance-based classification exhibits robustness across heterogeneous image sources. However, the study is limited to a single heritage object and a restricted number of devices, which may affect generalizability. The proposed framework provides a reproducible and interpretable method that supports digital authenticity verification and aligns with UNESCO’s vision for open, transparent cultural heritage preservation.
Volume: 15
Issue: 3
Page: 995-1003
Publish at: 2026-09-01
1 2 3
...
2097 2098
Show 23 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