Articles

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

Advanced Search
Filter Icon

Filters article

Years

FAQ Arrow
Single Year
Range Year
0
0

Source Title

FAQ Arrow

Authors

FAQ Arrow

31,377 Article Results

Exploring artificial intelligence in vocational learning: teachers’ perspective from Indonesia

10.11591/ijai.v15.i3.pp2009-2023
Yuliansah Yuliansah , Mar’atus Sholikah , Sutirman Sutirman
The rapid development of artificial intelligence (AI) in education has expanded its potential applications. However, empirical evidence from vocational education in developing countries remains limited, particularly regarding differences between certified and uncertified vocational high school (VHS) teachers’ perspectives. This study investigates VHS teachers’ perceptions of AI use in the learning process by explicitly comparing certified and uncertified teachers in Indonesia. Using a quantitative approach combined with data-mining techniques applied to open-ended survey responses, data were collected from 65 VHS teachers in the Special Region of Yogyakarta (DIY). Group differences were examined using a non parametric Mann-Whitney U test. The findings indicate that both certified and uncertified teachers demonstrate consistently positive perceptions of AI in instructional planning, implementation, and assessment, with no statistically significant differences between the two groups. Importantly, this result suggests that openness toward AI integration is not determined by certification status but reflects broader pedagogical orientations among vocational teachers. Teachers perceive AI primarily as a pedagogical partner rather than a substitute for professional educators. The study underscores the need of structured AI-focused professional development and policy support through adequate infrastructure and targeted training to enhance the effectiveness of AI adoption in improving the quality of vocational education in Indonesia.
Volume: 15
Issue: 3
Page: 2009-2023
Publish at: 2026-06-01

Performance comparison of deep learning models for concrete crack detection on mobile devices

10.11591/ijai.v15.i3.pp2811-2825
Sarapee Chunkaew , Somporn Ruang-On , Prawit Nuengmatcha , Kritaphat Songsri-in
Concrete crack detection is essential for structural maintenance, yet traditional manual inspection methods are time-consuming and require specialized expertise. While deep learning offers promising solutions, existing models often demand high computational resources unsuitable for mobile deployment. This research evaluates three convolutional neural network (CNN) architectures, namely mobile network (MobileNet), visual geometry group-16 (VGG-16), and residual network-50 (ResNet-50), to identify an optimal model for practical mobile-based crack detection. A dataset of 1,634 images was collected from online databases and field documentation, categorized into 10 classes across three severity levels: i) severe cracks requiring urgent repair (30%); ii) cracks requiring monitoring (40%); and iii) minor cracks (30%). The models were trained using standardized parameters with 224×224-pixel RGB input, rectified linear unit (ReLU) activation, and softmax classification. Systematic parameter optimization was conducted across epochs, learning rate, dropout rate, and optimizer selection, with stochastic gradient descent (SGD) identified as the optimal optimizer. Experimental results demonstrate that MobileNet achieves the best performance with 80% accuracy and a compact model size of 13.1 megabytes. This study concludes that MobileNet provides an optimal balance between detection accuracy and computational efficiency, enabling practical field deployment for automated concrete crack detection, with expert verification recommended for critical structural assessments.
Volume: 15
Issue: 3
Page: 2811-2825
Publish at: 2026-06-01

Detection of autism spectrum disorder using multilayer perceptron classifier

10.11591/ijai.v15.i3.pp2103-2112
Ahmed Q. Hadi , Saif H. Alrubaee , Fahad Taha Al-Dhief , Ammar AbdRaba Sakran
Autism spectrum disorder (ASD) is a neurodevelopmental condition marked by distinctive challenges in both verbal and nonverbal communication, social interaction, and repetitive behaviors. However, the diagnosis of ASD usually occurs within a clinical setting, conducted by licensed professionals, and often involves lengthy and costly procedures. On the other hand, machine learning holds significant promise for improving diagnostic and intervention research within the behavioral sciences, particularly in research concerning ASD disease. Hence, a deep investigation of a machine learning algorithm for ASD detection is crucial. Therefore, this paper presented a new system for differentiating the ASD samples from non-ASD (i.e., healthy) samples. The samples of ASD have been compiled from toddlers. The multilayer perceptron (MLP) algorithm is used to classify ASD samples from non-ASD samples. The proposed MLP classifier is implemented based on different numbers of neurons (i.e., nodes). In other words, the proposed MLP classifier started with 10 neurons and finished with 50 neurons with an increment step of 5 neurons. The outcomes demonstrate that the MLP classifier acquired different results concerning the number of neurons. The MLP obtained the best performance, reaching an accuracy rate of 100% in identifying ASD cases.
Volume: 15
Issue: 3
Page: 2103-2112
Publish at: 2026-06-01

Hyperparameter optimization of deep residual recurrent fusion models for facial emotion recognition

10.11591/ijai.v15.i3.pp2581-2594
Muhammad Munsarif , Ku Ruhana Ku-Mahamud
Deep learning facial emotion recognition (FER) is widely applied in healthcare, education, and human–computer interaction. However, many deep learning models suffer from suboptimal hyperparameter configurations that reduce accuracy and stability. This study proposes three deep residual recurrent fusion models that integrates residual blocks with recurrent neural networks (bidirectional long short-term memory (BiLSTM), long short-term memory (LSTM), and gated recurrent unit (GRU)) to capture both spatial and temporal features. A systematic hyperparameter optimization strategy was applied, tuning kernel size, filter size, recurrent units, batch size, learning rate, dropout, and weight decay to balance generalization and computational efficiency. The models were evaluated on four benchmark datasets: FER2013, FERPlus, RAF-DB, and CK+. The results show that optimized configurations achieved outstanding accuracy, reaching 99.85% on FER2013, 99.99% on FERPlus, and 100% on RAF-DB and CK+. These findings demonstrate that careful hyperparameter tuning significantly enhances feature extraction, mitigates vanishing gradient and overfitting issues, and improves generalization across diverse datasets. The proposed framework highlights the importance of optimization in advancing robust FER systems for real-world applications.
Volume: 15
Issue: 3
Page: 2581-2594
Publish at: 2026-06-01

A hybrid deep learning approach for BoT-IoT intrusion detection

10.11591/ijai.v15.i3.pp2192-2200
Khalid Altarawneh , Ghayth AlMahadin , Ibrahim Altarawni
Internet of things (IoT) devices enhance quality of life and industrial operations but pose significant security risks, necessitating intelligent intrusion detection systems (IDS) to combat evolving cyber threats. This paper proposes a novel IDS framework integrating bio-inspired heuristic feature selection, a generative adversarial network (GAN)-based data augmentation, and an ensemble classifier combining ResNet, AlexNet, and MobileNet. The methodology, tested on the botnet (BoT)-IoT dataset, follows four stages: preprocessing, feature augmentation, feature selection, and ensemble classification. Evaluated on benchmarks including CIC-IDS-2018, NSL-KDD, and UNSW-NB15, the model achieved accuracies of 98.2%, 99.1%, 97.6%, and 98.4%, respectively, with consistently high precision, recall, and F1-scores, demonstrating robust detection of diverse cyberattacks. Beyond accuracy, the framework optimizes processing time for large-scale IoT data, addressing scalability challenges in real-time threat mitigation. By synergizing feature optimization, synthetic data generation, and deep learning architectures, the solution enhances detection rates while minimizing computational overhead. Comparative analysis highlights its superior performance over existing methods, positioning it as a vital tool for securing IoT ecosystems against unauthorized access and malicious activities. The results underscore its potential to fortify IoT network security, balancing efficiency, adaptability, and computational feasibility for practical deployment in resource-constrained environments.
Volume: 15
Issue: 3
Page: 2192-2200
Publish at: 2026-06-01

Enhancing fake news detection: a hybrid BERT-XGBoost model for improved performance and interpretability

10.11591/ijai.v15.i3.pp2385-2397
Nishant Vasantkumar Hegde , Suneesh Bare , Namruth Reddy , Rajat Gondkar Aravinda , Minal Moharir , Aamir Ibrahim
The widespread spread of fake news poses a serious threat to the integrity of information. The dominant approach to detection involves end-to-end fine-tuning of large transformer models like bidirectional encoder representations from transformers (BERT), which, despite achieving high accuracy, often function as opaque “black boxes” with limited interpretability. This paper proposes and validates a hybrid, decoupled architecture that proves to be a more practical and powerful alternative. We first fine-tune a DistilBERT model on the full WELFake dataset of 71,537 articles after cleaning to create domain-specific embeddings. These high-dimensional vectors are then used as input features to train a robust extreme gradient boosting (XGBoost) classifier. The results demonstrate that the hybrid model achieves a state-of-the-art accuracy of 99.76%, slightly surpassing the already high performance of a standard end-to-end fine-tuned model. Crucially, this approach provides this top-tier performance while offering significant advantages in model interpretability through feature importance analysis. This work establishes that a decoupled architecture is not just a viable alternative but a superior practical strategy for combating misinformation, successfully balancing state-of-the-art accuracy with essential model transparency.
Volume: 15
Issue: 3
Page: 2385-2397
Publish at: 2026-06-01

Transformer-based hybrid classification for plant leaf disease detection using vision transformer, principal component analysis, and support vector machine

10.11591/ijece.v16i3.pp1399-1406
Vijayalakshmi S. Abbigeri , Geetha D. Devanagavi
Plant diseases remain a critical challenge in agriculture, causing substantial yield losses and threatening food security. In this work, we propose a hybrid deep feature engineering framework that integrates deep learning-based feature extraction with classical machine learning for accurate plant disease detection. A pretrained vision transformer (ViT) model is employed to extract discriminative features from leaf images, effectively capturing complex spatial relationships. To address the curse of dimensionality, principal component analysis (PCA) is applied, retaining 98% of the variance while reducing feature space complexity. The refined features are then classified using a support vector machine (SVM) optimized through hyperparameter tuning. Experimental results on the bean leaf lesions dataset demonstrate strong performance, achieving 92% accuracy and a weighted F1-score of 0.92. The proposed ViT–PCA–SVM pipeline effectively balances accuracy, computational efficiency, and generalization, making it a promising solution for real-time smart farming applications.
Volume: 16
Issue: 3
Page: 1399-1406
Publish at: 2026-06-01

Ensuring effective cervical cancer diseases diagnosis system using ensemble machine learning model

10.11591/ijai.v15.i3.pp2410-2422
Oluwatobi Akinlade , Jafar Abdollahi , Misan Paul Etchie , Sunday Adeola Ajagbe , Oluwaseyi Omotayo Alabi , Bambo Ayo Adeyanju
Worldwide, cervical cancer is a major public health concern, and early detection is essential for successful management and treatment. Ensemble machine learning (ML) has emerged as a promising approach for improving the accuracy and reliability of cervical cancer diagnosis systems. This study evaluated the performance of various ensemble ML models for cervical cancer diagnosis using a large dataset of cervical cell images. The performance of different ensemble models was compared, including random forest (RF), gradient boosting (GB), and stacking, with conventional ML models, such as logistic regression and support vector machine (SVM). The bagging ensemble model developed reported training of 0.991667, 0.827586, 1.0, and 0.90566 for accuracy, precision, recall, and F-measure (F1-score), respectively. Furthermore, the interpretability of ensemble models was investigated using feature importance and partial dependence plots. The interpretability analysis revealed that ensemble models can provide valuable insights into the key features and factors that contribute to cervical cancer diagnosis. In conclusion, the findings suggest that ensemble ML is a promising approach for developing accurate and reliable cervical cancer diagnosis systems and improving the interpretability of these models.
Volume: 15
Issue: 3
Page: 2410-2422
Publish at: 2026-06-01

ValveHealthNet: a light deep learning model for accurate valvular heart disorder detection

10.11591/ijai.v15.i3.pp2643-2654
Ausilah Alfraihat , Wafaa Al-Sharu , Ali Mohammad Alqudah
Valvular heart disease (VHD) is a significant global health issue, contributing to increased morbidity and mortality rates, particularly in aging populations. Current diagnostic methods, such as echocardiography and manual auscultation, face limitations in accessibility and accuracy, particularly in resource-constrained environments. This study introduces ValveHealthNet, a lightweight deep learning model designed to classify various VHDs using heart sound recordings. Leveraging a dataset of over 10,000 heart sounds, minimal preprocessing was applied by converting the audio signals into power spectra before feeding them into a convolutional neural network (CNN) combined with a bidirectional long short-term memory (BiLSTM) network. This model achieved impressive results, with an accuracy of 98% in training and testing and 98.4% through 10-fold cross-validation. This highly efficient model can be used in embedded systems, providing a cost-effective, AI-driven solution for early detection of VHD in settings where advanced diagnostic tools may be unavailable.
Volume: 15
Issue: 3
Page: 2643-2654
Publish at: 2026-06-01

Identification of areas of influence using a thematic modeling approach and belief function theory

10.11591/ijai.v15.i3.pp2838-2848
Fatima-Zahrae Sifi , Wafae Sabbar , Amal El Mzabi
Social networks have become essential platforms for the dissemination of information and the exchange of ideas. They are transforming the way influencers interact and exert influence over their peers. On platforms such as Twitter, Facebook and Instagram, influence is expressed through social interactions such as retweets, likes, mentions, comments and shares. These activities play a key role in amplifying messages and shaping opinions. Studying the practices of influencers allows for a more precise identification of the domains in which their impact is particularly significant. However, this task is complex. It requires rigorous methods capable of integrating various forms of social engagement while managing the uncertainties associated with heterogeneous data. In this context, we propose a method to identify the domain of influence of a social media influencer. This approach combines thematic modeling Latent Dirichlet Allocation (LDA) with Belief Function Theory (BFT) to analyze social interactions and dominant topics of interest. By incorporating indicators such as retweets, likes and mentions, the method provides a robust framework for evaluating the influencer's impact across different domains. It thus offers precise tools for researchers, practitioners and decision-makers aiming to better understand these complex dynamics.
Volume: 15
Issue: 3
Page: 2838-2848
Publish at: 2026-06-01

Prediction of new student admissions to higher education using support vector machines

10.11591/ijai.v15.i3.pp2484-2493
Neni Purwati , Windya Harieska Pramujati , A. Aviv Mahmudi , Mira Febriana Sesunan , Yahya Yahya
Higher education institutions across various regions operate using systems that generate large amounts of data. This data is stored and utilized for strategic decision-making, providing significant business value to these institutions. Support vector machine (SVM) has become popular due to its strong generalization capability, high prediction accuracy, and faster training speed. SVM employs kernels as tuning parameters. This study aims to enhance the accuracy of student admissions prediction in higher education institutions using the SVM classification model. The SVM model was applied to a dataset comprising 5,936 records with four attributes and was evaluated using the use training set, 10-fold cross-validation, and percentage splits of 70%–30% and 80%–20%. Initially, the SVM-kernel model achieved high accuracy but failed to identify any true positive instances, indicating its inability to detect the minority “not accepted” class due to severe class imbalance. After applying class balancing techniques, the model’s performance improved significantly in terms of area under the curve (AUC), F-measure, and Matthews correlation coefficient (MCC), reflecting a more balanced classification between majority and minority classes. The SVM with Pearson VII function-based universal kernel (PUK) and classifier version 4.5 (C4.5) models achieved the best performance, indicating that class balancing effectively enhances both sensitivity and fairness in predictive classification.
Volume: 15
Issue: 3
Page: 2484-2493
Publish at: 2026-06-01

Optimized classification of student performance outcomes using LEE feature selection in the context of educational data mining

10.11591/ijai.v15.i3.pp2459-2470
Kishore Kumar Kamarajugadda , Movva Pavani , Rani Vanathi Gurusamy , Nagarajan Karthikeyan , Pavan Kumar Nidumolu , Desidi Narsimha Reddy , Muniappan Ramaraj , Rajasekaran Nithya
Student speculative victory is a vital area that needs to be predicted to improve the quality of education and aid the institutional decision making. This research work has to planned to use learning based enhanced evaluation (LEE) feature selection method with real world educational datasets for optimized data mining approach to predict student performance. High dimensionality and irrelevant features are common problems with enhanced models, affecting classification accuracy and efficiency. LEE feature algorithm is used to extract important features, that enhance the performance of the model, reduce the calculation quantity of the model. The methodology consists of pre-processing of the dataset, feature selection using LEE algorithm, and testing four classifiers namely support vector machine (SVM), k-nearest neighbor (KNN), adaptive learning, and naïve Bayes. The incorporation of LEE improves the model’s ability by reducing noise and highlighting the influential features. Experimental results show that optimized techniques are better in terms of accuracy and robustness than others. The models are evaluated based on important performance metrics such as accuracy, precision, recall, F1-score, and training time. The enhanced approach will help to add to the literature of the field of educational data mining (EDM), providing a practical and effective way of predicting student performance in real academic settings.
Volume: 15
Issue: 3
Page: 2459-2470
Publish at: 2026-06-01

Cutting-edge algorithms for reliable failure prediction in metro train systems

10.11591/ijai.v15.i3.pp2269-2277
Sana Chakri , Naoual Mouhni , Faouzia Ennaama
This study investigated various machine learning algorithms on dataset for failure prediction within metro train systems. The data indicated strong linear relationships within the dataset, making linear models such as support vector machines (SVMs) viable, as well as logistic regression analysis. For example, the least absolute shrinkage and selection operator (LASSO) regularization method used in feature selection had profound implications, leading to enhanced performance through the identification of pertinent attributes. Some advanced models like gradient boosting machines (GBMs), convolutional neural networks (CNNs), and kernel SVMs were found to outperform the conventional methods because they are capable of recognizing any complicated trends or non-linear relationships present in data sets. Combining strong learners can produce an ensemble model that improves forecast performance, while top-performing models are used in the ensemble method to enhance prediction accuracy. These findings would help professionals in the metro train industry choose appropriate machine learning methods to support preventive maintenance strategies, minimizing costs while enhancing operational effectiveness and safety.
Volume: 15
Issue: 3
Page: 2269-2277
Publish at: 2026-06-01

Vision transformer and hybrid models for Malayalam handwritten word recognition

10.11591/ijai.v15.i3.pp2655-2663
Anju Arangil Thazhath , Binu Poothakuzhiyil Chacko , Mohamed Basheer Kizhakke Parambath
Transformer-based architectures and attention mechanisms have revolutionized the field of image recognition. This study focuses on offline handwritten Malayalam word recognition, addressing the lack of publicly available datasets for this low-resource language. A new Malayalam word dataset (MWD) comprising 20,850 samples across 139 classes was developed to support research in this domain. The vision transformer (ViT) was employed for advanced feature extraction, and multiple recognition models—feed-forward neural network (FFNN), global average pooling (GAP), bidirectional long short-term memory (BiLSTM), and attention based feed-forward neural network (AFFNN)—were evaluated. Among these, AFFNN achieved the highest accuracy of 98.56%, establishing the proposed vision transformer-based attention handwritten word recognition (ViTA-HWR) model as a robust framework for handwritten Malayalam word recognition and valuable contribution to regional language processing.
Volume: 15
Issue: 3
Page: 2655-2663
Publish at: 2026-06-01

Architectural design of an internet of things-based framework for road bike speed optimization

10.11591/ijai.v15.i3.pp2125-2140
Tigor Hamonangan Nasution , Opim Salim Sitompul , Fahmi Fahmi , Muhammad Anggia Muchtar
This research aims to develop an internet of things (IoT) system framework to predict cyclists’ optimal speed in road cycling using multisensor data and machine learning. The primary issue raised is the lack of an intelligent system capable of integrating physiological, performance, and environmental data in real-time speeds for cyclists. The designed framework consists of four functional layers: data acquisition layer; data processing and feature layer; predictive modeling layer; and recommendations and output layer. Modeling is carried out using gradient boosting regression (GBR), performed end-to-end with validation on real cyclist activity data. The test results demonstrate that the system can provide precise optimal speed estimates and offer pacing zone recommendations that positively impact athlete performance strategies. This research contributes novelty in the form of an adaptive multivariate prediction approach and a modular IoT architecture design that can be implemented on cloud and edge platforms.
Volume: 15
Issue: 3
Page: 2125-2140
Publish at: 2026-06-01
1 2 3
...
2091 2092
Show 65 of 2092

Discover Our Library

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

Explore Now
Library 3D Ilustration