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

29,922 Article Results

An efficient method to improve machine learning decoders using automorphisms group

10.11591/ijai.v15.i1.pp547-558
Imrane Chemseddine Idrissi , Said Nouh , El Mehdi Bellfkih , Mohammed El Assad , Abdelaziz Marzak
The decoding of error-correcting codes (ECCs) is a critical aspect of communication systems, yet traditional decoding techniques can often be computationally demanding or ineffective for certain codes, necessitating innovative approaches. In this study, we introduce a hybrid approach that combines machine learning and automorphism techniques to optimize the decoding process. Specifically, we train multilayer perceptron (MLP) models to learn the mapping between error syndromes and their corresponding errors. While these models exhibit robust learning capabilities, their performance sometimes does not reach 100%. To mitigate this limitation, we exploit the automorphism group of the code—a set of structure-preserving transformations—to convert the errors that the MLP struggles to decode into ones it can process more effectively. We use a minimum number of p permutations, pre-calculating and storing all possible automorphisms to ensure computational efficiency. Our experimental results reveal that this hybrid approach substantially enhances the decoding performance of the MLP model, presenting a promising avenue for decoding ECCs. Importantly, this approach is not limited to MLP models and can be applied to any machine learning model with a learning score less than 100%, broadening its applicability and impact. By integrating machine learning with traditional algebraic coding theory, we propose a new paradigm that holds the potential to revolutionize the design of decoding systems, making them more efficient and effective.
Volume: 15
Issue: 1
Page: 547-558
Publish at: 2026-02-01

Instagram influencer classification using fine-tuned BERT model

10.11591/ijai.v15.i1.pp1009-1018
Ni Putu Sutramiani , Ni Made Dita Dwikasari , I Nyoman Prayana Trisna , I Wayan Agus Surya Darma
Influencer marketing has emerged as a powerful strategy in today’s digital world, where social media stars can influence how people think about products. However, the rapid growth of influencers and social media users presents novel challenges for brands in identifying suitable influencers for their marketing goals. Traditional approaches that rely on popularity and follower count are no longer the primary metrics for determining an influencer’s ability to affect consumer behavior. To address this gap, this study proposed an influencer classification to enhance audience targeting and marketing effectiveness. By utilizing deep learning, specifically fine tuned bidirectional encoder representations from transformers (BERT), influencer classification was carried out for Instagram users in Indonesia based on their post captions. The multilingual BERT model is optimized through hyperparameter tuning, including learning rate, batch size, and stop word removal variation. With an outstanding 80% accuracy, the model performs best in situations where stop words are not removed. This study on influencer classification using a fine-tuned BERT model has demonstrated the effectiveness of BERT in enhancing influencer selection. It contributes to the digital marketing domain by showcasing the potential of deep learning for social media analysis and content classification, paving the way for future data-driven marketing strategies.
Volume: 15
Issue: 1
Page: 1009-1018
Publish at: 2026-02-01

Malware detection using convolutional neural network-di strategy polar fox optimization algorithm

10.11591/ijai.v15.i1.pp140-153
Parvathi Sathenahalli Jayaprakash , Yogeesh Ambalagere Chandrashekaraiah
Malware attacks have escalated significantly with an increase of internet users and connected devices. With the rise of various types of malwares released by the hackers, constructing new competitive methods are necessary to identify the advanced malware. However, conventional malware detection struggles to identify new and evolving malware variants accurately because of its dependence on handcrafted features and static-signature based methods. To address this problem, this research proposes convolutional neural network (CNN) based di strategy polar fox optimization algorithm (DSPFOA) for malware detection to fine-tune the CNN parameters effectively which later assists to overcome the limitations of CNN. The model integrates the sine chaotic mapping and Cauchy operator mutation as DSPFOA prevents the model from local optima issue, and extends search space solution, also enhance convergence. This ensures that the CNN learns highly discriminative features which makes the system more accurate and robust in detecting both known and evolving malware variants. The CNN DSPFOA achieves a high accuracy of 99.65 and 99.76% by utilizing BIG2015 and Malimg dataset respectively compared to existing methods like masked self-supervised model with swin transformer (MalSort).
Volume: 15
Issue: 1
Page: 140-153
Publish at: 2026-02-01

A comprehensive survey of cyberbullying on social media: challenges, detection, and AI-based prevention

10.11591/ijai.v15.i1.pp86-96
Ammar Odeh , Osama Alhaj Hassan , Anas Abu Taleb , Abobakr Aboshgifa , Nabil Belhaj
Cyberbullying is a pervasive issue in the digital landscape, particularly on social media platforms, where individuals engage in online harassment, intimidation, and abuse. Unlike traditional bullying, cyberbullying has a broader reach, anonymity, and persistence, making it a growing concern for mental health, social well-being, and online safety. This paper provides a comprehensive survey of cyberbullying trends, its psychological and social impacts, and the role of social media in amplifying the problem. It explores existing detection and prevention strategies, including artificial intelligence (AI)-driven approaches, policy frameworks, and platform-based moderation techniques. Furthermore, it discusses challenges in enforcement, the limitations of automated detection systems, and the need for improved legal measures. This paper uniquely contributes an integrated perspective on cyberbullying detection and prevention by synthesizing current research across psychological, sociocultural, and technical dimensions. It emphasizes underexplored gaps such as multilingual detection, real-time moderation, and cross-platform enforcement, and proposes a layered framework to guide future research and policy.
Volume: 15
Issue: 1
Page: 86-96
Publish at: 2026-02-01

Optimization of maximum power point tracking in wind energy systems: a comparative study of ant colony and genetic algorithms

10.11591/ijai.v15.i1.pp399-411
Najoua Mrabet , Chirine Benzazah , Mohssine Chakib , Adil Ziraoui , Ahmed El Akkary , Najma Laaroussi
This research focuses on optimizing maximum power point tracking (MPPT) in wind energy conversion systems (WECS) using ant colony optimization (ACO) and genetic algorithm (GA). The study evaluates these two metaheuristic techniques to optimize the parameters of a proportional integral-derivative (PID) controller in order to maximize power output in a permanent magnet synchronous generator (PMSG)-based system. Simulations conducted in MATLAB/Simulink show that both ACO and GA effectively enhance MPPT performance by improving power output, DC bus voltage regulation, and torque stability. The results demonstrate the potential of metaheuristic algorithms to optimize wind energy conversion efficiency and support sustainable energy development.
Volume: 15
Issue: 1
Page: 399-411
Publish at: 2026-02-01

Predicting trapped victims in debris using signal analysis ensemble classification

10.11591/ijai.v15.i1.pp493-505
Enoch Adama Jiya , Ilesanmi B. Oluwafemi , Olayinka O. Ogundile , Oluwaseyi P. Babalola
One major difficulty in pervasive computing is trapped human detection in search and rescue (SAR) scenarios. Accurately identifying trapped individuals is challenging due to noisy data and the curse of dimensionality. When non-line-of-sight (NLOS) conditions are present during catastrophic occurrences, the curse of dimensionality can result in blind spots in detections because of noise and uncorrelated data. Because machine learning algorithms are incredibly accurate, this work focuses on using ultra wideband (UWB) radar waves to detect individuals in NLOS scenarios and leveraging wireless communication to harmonize information. The paper uses ensemble methods to extract features using independent component analysis (ICA) and evaluate classification performance on both static and dynamic datasets. The testing results confirm the effectiveness of the proposed strategy, with classification accuracies of 87.20% for dynamic data and 88.00% for static data. Lastly, during SAR operations, our approach can assist engineers and scientists in making quick decisions.
Volume: 15
Issue: 1
Page: 493-505
Publish at: 2026-02-01

An artificial intelligence technology for promoting hom-thong banana agriculture system

10.11591/ijai.v15.i1.pp568-579
Ratsames Tanveenukool , Suwit Somsuphaprungyos , Boonyarit Nokkurth , Likit Chamuthai , Patumwadee Bonguleaum , Parinya Natho
The hom-thong banana, being a high-value Thai export variety, is facing significant risk from disease outbreaks affecting crop yield and quality. Traditional visual inspection methods in detection of diseases are labor consuming, error-prone. This research addresses these limitations by developing a new artificial intelligence (AI)-based automatic disease detection system for the hom-thong banana industry on top of cutting-edge computer vision technology. The study employed deep learning object detection models, contrasting Roboflow, you only look once (YOLO)v11, and YOLOv12 architectures, which were trained on a large dataset of 2,576 images of Thai banana plantations. With systematic data augmentation techniques, the dataset was augmented to 6,184 images of seven types of disease under varied environmental conditions. The method entailed extensive preprocessing and evaluation of performance through precision, recall, and mean average precision (mAP) metrics. Outcomes indicated that YOLOv12 outperformed with 93.3% accuracy, 83.3% sensitivity, and 86.3% mAP@50 compared to standard inspection schemes. This research is applicable to Thailand's smart agriculture initiative by providing farmers with low-cost, accurate, and effective disease monitoring equipment. The application of this AI system has the ability to enhance the yield of crops, reduce losses, and enhance the competitiveness of Thai banana exports in the global market, in support of sustainable agricultural development.
Volume: 15
Issue: 1
Page: 568-579
Publish at: 2026-02-01

Multi-scale features assisted knowledge distillation vision transformer for land cover segmentation and classification

10.11591/ijai.v15.i1.pp361-373
Sujata Arjun Gaikwad , Vijaya Musande
The most significant problem in remote sensing interpretation is semantic segmentation, which attempts to give each pixel in the image a particular class. This research work follows the various steps, such as pre-processing, segmentation, and classification. Initially, high spatial resolution remote sensing images (RSI) are collected from the open-source dataset. In the pre processing stage, an improved guided filter (Imp-GF) is used to remove various noises from images. Next, the segmentation is done by using a knowledge distillation-based vision transformer approach integrated with an atrous spatial multi-scale pyramidal module (KD-MuViTPy). Based on the segmented image, land cover classes such as vegetation, urban areas, forest, water bodies, and roads are classified. The proposed method outperformed the Bhuvan satellite dataset, achieving better accuracy, precision, recall, F1 score, Dice score, intersection over union (IoU), and Kappa score at values of 98.01%, 98.99%, 97.49%, 98.23%, 98.23%, 96.55%, and 95.91%, respectively.
Volume: 15
Issue: 1
Page: 361-373
Publish at: 2026-02-01

Scalable resume screening using large language model Meta AI version 3

10.11591/ijai.v15.i1.pp953-961
Asmita Deshmukh , Anjali Raut , Vedant Deshmukh
This research paper explores the use of large language model Meta AI 3 (LLaMA 3) for automating the resume screening process. Traditional resume screening methods that rely on keyword searching and human review can be inefficient, biased, and fail to identify qualified candidates. LLaMA 3, trained on large-scale text datasets, has the potential to accurately analyze resumes by understanding context and semantic details beyond simple keyword matching.The study presents a system that converts resume PDFs to text, inputs the text along with the job description into the LLaMA 3 model, and generates a ranked list of candidates with reasoning for their job fit. This discusses the data preparation, model setup, and performance evaluation of this system. Results show LLaMA 3 can rapidly process batches of resumes while reducing human bias in the screening process. The system aims to streamline hiring by automating the initial resume screening stage to surface top candidates for further in-depth evaluation. Key benefits include improved accuracy in identifying relevant skills, reduced bias compared to human screeners, and significant time savings for recruiters. The paper also examines ethical considerations around using AI for hiring decisions. Overall, this work demonstrates the promising application of large language models (LLMs) like LLaMA 3 to transform and enhance resume screening practices.
Volume: 15
Issue: 1
Page: 953-961
Publish at: 2026-02-01

Stroke prediction using data balancing method and extreme gradient boosting

10.11591/ijai.v15.i1.pp655-671
Abd Mizwar A. Rahim , Anna Baita , Firman Asharudin , Wahid Miftahul Ashari , Walidy Rahman Hakim , Andriyan Dwi Putra , Supriatin Supriatin , Eko Pramono
Stroke is one of the leading causes of death worldwide, creating an urgent need for effective early detection systems, particularly because conventional methods often struggle with class imbalance and produce biased evaluations. Previous studies have primarily focused on accuracy while overlooking model consistency, data pre-processing quality, and probability-based evaluation. This study evaluates model performance under three conditions: original data using extreme gradient boosting (XGBoost) with scale_pos_weight, original data using the easy ensemble classifier, and class-balanced data generated using random oversampling (ROS), adaptive synthetic sampling (ADASYN), and synthetic minority over-sampling technique (SMOTE). Each model underwent missing value handling, normalization, feature preparation, and hyperparameter optimization using grid search. Performance was assessed using area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), confidence intervals, calibration curves, Shapley additive explanations (SHAP), decision curve analysis (DCA), and external validation. The results demonstrate that data resampling significantly improves performance, with the XGBoost-SMOTE combination achieving the best results, including an accuracy of 0.99, AUROC of 0.998, and AUPRC of 0.986, outperforming the other approaches. This method provides more consistent and balanced predictions, supporting the application of artificial intelligence for early stroke risk identification.
Volume: 15
Issue: 1
Page: 655-671
Publish at: 2026-02-01

Automated ergonomic sitting postures detection for office workstation using XGBoost method

10.11591/ijai.v15.i1.pp506-514
Theresia Amelia Pawitra , Farida Djumiati Sitania , Anindita Septiarini , Hamdani Hamdani
Sedentary office work increases musculoskeletal risk, underscoring the need for non-intrusive, real-time posture monitoring. This study presents a computer vision approach that classifies ergonomic versus non-ergonomic sitting postures using upper body key points extracted by MoveNet thunder. Images from 30 participants were captured from frontal and side views, and labeled according to SNI 9011:2021 criteria. Seventeen key points were detected, with head-to-hip landmarks retained, then normalized and centered. Three classifiers—adaptive boosting (AdaBoost), extreme gradient boosting (XGBoost), and a multi-layer perceptron (MLP)—were trained and evaluated with 10-fold stratified cross-validation. XGBoost achieved the best performance, with accuracy 93.0%±1.9%, precision 94.6%, recall 91.4%, F1-score 92.9%, and area under the receiver operating characteristic curve (ROC-AUC) 0.974±0.010, outperforming MLP and AdaBoost. The method supports privacy-preserving, on-device inference and is suitable for integration into smart office systems to reduce exposure to high-risk postures. Limitations include controlled capture conditions and an upper body focus; future work will expand posture taxonomy and real-world deployment.
Volume: 15
Issue: 1
Page: 506-514
Publish at: 2026-02-01

Anefficient ensemble tree-based framework for intrusion detection in industrial internet of things networks

10.11591/ijai.v15.i1.pp481-492
Mouad Choukhairi , Oumaima Chentoufi , Ouail Choukhairi , Youssef Fakhri
The increasing complexity of cyber threats in industrial internet of things (IIoT) environments necessitates robust, scalable, and efficient intrusion detection systems (IDS). This study presents a novel ensemble tree-based framework that integrates gradient boosting-based machine learning models, including XGBoost, LightGBM, AdaBoost, and CatBoost, with mutual information (MI) feature selection and synthetic minority over-sampling technique (SMOTE) to enhance multiclass intrusion detection performance. The framework is designed to handle large-scale, imbalanced datasets efficiently while maintaining high classification accuracy. Performance evaluation using the telemetry of network (ToN)-IoT benchmark dataset demonstrates that the proposed models achieve a high accuracy of 99.43%, with a strong precision-recall balance and an F1-score, ensuring minimal false positive rates of 0.08%. By leveraging MI for optimal feature selection and SMOTE for data balancing, this approach effectively enhances detection capabilities in highly dynamic network environments. The lightweight architecture and reduced execution time make the framework well-suited for deployment in edge or fog nodes within smart industrial environments. The proposed solution provides a scalable and adaptable methodology for securing IIoT networks, making it applicable for real-time intrusion monitoring and further cybersecurity advancements in industrial systems.
Volume: 15
Issue: 1
Page: 481-492
Publish at: 2026-02-01

Adversarial examples in Arabic language

10.11591/ijai.v15.i1.pp941-952
Safae Laatyaoui , Mohammed Saber
Adversarial attacks have a great popularity in the artificial intelligence (AI) domain. In the natural language processing (NLP) field, various techniques have been used to evaluate the vulnerability of deep learning (DL) models. It is observed that while most studies focused on generating adversarial examples in English language, Arabic adversarial attacks have received little attention. This paper presents a two-step method to create adversarial examples in Arabic language: first, the most important words are identified. Then, the proposed transformation algorithm is applied. Only small and imperceptible manipulations based on common mistakes in Arabic writing mislead the popular pre-trained language model (PLM) bidirectional encoder representations from transformers (BERT) retrained on the book reviews in Arabic dataset (BRAD) on the sentiment analysis (SA) task and decrease its performance: the classification accuracy was reduced by an average of 3.44%. This drop in accuracy shows that the model was successfully attacked.
Volume: 15
Issue: 1
Page: 941-952
Publish at: 2026-02-01

Adaptive transformer architecture for scalable earth observation via hyperspectral imaging

10.11591/ijai.v15.i1.pp824-830
Devendra Kumar Saragoor Madanayaka , Devanathan Muthukrishnan
Hyperspectral Image (HSI) classification is one of the critical processes involved in remote sensing application that plays a crucial role towards earth observation. Owing to complex spatial-spectral relationship and high dimensionality, it is quite a challenging task to subject HSI content to conventional data analytics or existing methods. Hence, the proposed study introduces a novel computational model known as Adaptive Spectra-Spatial Transformer (ASST) to address these ongoing challenges and shortcoming of existing Artificial Intelligence (AI) based modelling. The proposed model contributes towards a novel transformer-based architecture where a distinct spectral-spatial attention method has been used with transformer encoder. This novel combination facilitates highly adaptive and contextually enriched feature extraction. Tested on universally standard HSI dataset of Pavia University, the proposed ASST model has been benchmarked with notice 97.26% of overall accuracy and faster processing duration computed via training and response time in contrast to frequently adopted ML and DL models. The accomplished study outcomes truly exhibited highly improved feature representation as well as robust performance against class imbalance problems towards scalable data analysis of HSI contents for earth observation.
Volume: 15
Issue: 1
Page: 824-830
Publish at: 2026-02-01

Improving efficiency of autism detection based on facial image landmarks

10.11591/ijai.v15.i1.pp766-779
Nguyen Trong Tung , Ngo Duc Vinh , Ha Manh Toan , Do Nang Toan
Autism is a serious mental health problem with long-term effects on life. Therefore, early diagnosis is a topical issue for effective treatment. This study proposes a novel facial landmark transformation-based data augmentation method that allows for the generation of geometric transformations related to facial geometry. This method increases the generalizability and provides a perspective on the role of facial regions in autism detection. The proposed augmentation method ensures the generation of variants that are consistent with the facial image structure and the nature of the facial image. Next, conduct a comprehensive and comparative study with EfficientNet-B0, EfficientNet-B4, ResNet-18, ResNet-50, ResNet-101, MobileNet-V2, DenseNet-121 and DenseNet-201. Also analyze the model's attention over the main regions of the face that are related to facial landmarks. The results clearly show that the models trained with the proposed method outperform the default augmentation method. Specifically, when averaging the measures across the tested models, the results are 0.905417 for accuracy, 0.962133 for area under the curve (AUC), 0.9198 for precision, 0.888333 for recall, and 0.903678 for F1-score. Furthermore, when analyzing the gradient-weighted class activation mapping (Grad CAM) heatmaps, the high-value regions are clearly concentrated on the main areas of the face. Source code is published on GitLab platform.
Volume: 15
Issue: 1
Page: 766-779
Publish at: 2026-02-01
1 2 3
...
1994 1995
Show 46 of 1995

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