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

Text emotion recognition based on deep learning and attention with whale optimization algorithm

10.11591/ijai.v15.i4.pp3692-3702
Tabarek Alwan Tuib , Mohammad-Reza Feizi-Derakhshi , Yaqdhan Mahmood Hussein , Fahad Taha Al-Dhief
Text emotion recognition (TER) represents one of the crucial tasks in natural language processing (NLP), and it is highly important for many applications. To improve the performance of TER, advanced deep learning (DL) model GRU-BiGRU-CNN-Att is presented in this research. It is a combination of gated recurrent units (GRU), bidirectional gated recurrent units (BiGRU) with a convolutional neural network (CNN), and an attention mechanism. Using a GRU, BiGRU, and CNN as part of a deep feature extraction model, combined with an attention mechanism that gives important information different weights, the method that is proposed in the present work enhances the quality of the word vectors and leads to increasing sentiment analysis judgment accuracy. The whale optimization algorithm (WOA), which selects the most informative features for further enhancing the model, is used lastly for the optimization of the feature selection process. After such optimization, those chosen features are trained on a multi-layer perceptron (MLP), successfully combining machine learning (ML) and DL approaches for improving TER. A varied corpus of tweets, sentences, and dialogues had been used in the present work for the assessment of the performance of the suggested model. The proposed method achieves 83.76% accuracy in emotion recognition.
Volume: 15
Issue: 4
Page: 3692-3702
Publish at: 2026-08-01

Transforming correspondence learning through a smart assistant system to improve student outcomes and engagement

10.11591/ijai.v15.i4.pp3218-3227
Marsofiyati Marsofiyati , Roni Faslah , Maulana Amirul Adha , Muhammad Ikhwan , Faerozh Bin Madli
The insufficient feedback in business correspondence education at the higher education level necessitates innovation in adaptive and responsive learning resources. This study seeks to create and evaluate the efficacy of an artificial intelligence (AI)-driven correspondence learning assistant (CORLA) to enhance students' correspondence abilities. The study employs a research and development (R&D) methodology by utilizing the Dick and Carey model, which has ten phases of instructional development. The participants in the study were students enrolled in Digital Office Administration Study Program, Universitas Negeri Jakarta. Data were gathered by observation, interviews, questionnaires, and assessments (pre-test and post-test), and evaluated by three experts in materials, learning design, and media. The findings indicated that CORLA demonstrates a significant degree of feasibility, practicality, and efficacy. CORLA's advanced features encompass AI-driven automatic feedback, letter composition simulation, learning progress monitoring, discussion forums, and an intuitive user interface (UI). This study advances the creation of an effective and practical AI-based learning media model tailored to the requirements of teaching business correspondence abilities in higher education.
Volume: 15
Issue: 4
Page: 3218-3227
Publish at: 2026-08-01

YOLOv11 optimization for tiny object in crowded scenes

10.11591/ijai.v15.i4.pp3452-3463
Husna Sarirah Husin , Howard Chong Yun Hao , Pan Yuan Fei , Mohsen Marjani , Suriana Ismail
Small object detection in crowded urban and aerial scenes remains a critical challenge due to limited pixel information and information loss in deep neural networks. This study introduces a novel optimization framework for YOLOv11, specifically engineered for tiny-scale targets by integrating convolutional block attention modules (CBAM), k-means anchor clustering, and an enhanced feature pyramid network (FPN). Evaluated on the TinyPerson and COCO-mini datasets, the YOLOv11-optimized model achieves significant performance breakthroughs, delivering a +7.3% gain in mean average precision (mAP) and a +10.5% increase in recall over the baseline. Notably, the model achieved a recall of 0.072 on the TinyPerson dataset, with double sensitivity of standard YOLOv11. With a high-speed inference rate of 27.3 FPS, this research demonstrates that strategic architectural refinements can drastically improve small object detection reliability without compromising real-time viability on edge devices.
Volume: 15
Issue: 4
Page: 3452-3463
Publish at: 2026-08-01

Biometric authentication using dual-modal deep learning based-on electrocardiogram and ear features

10.11591/ijai.v15.i4.pp3252-3268
Mohamed S. Khalaf , Said Fathy Al-Zoghdy , Mariana Barsoum , Ibrahim Omara
Biometric authentication systems are essential for secure access control; however unimodal systems are vulnerable to spoofing and environmental variations. This study proposes a dual-modal biometric system combining electrocardiogram (ECG) signals and ear features to enhance security and accuracy. Deep learning architectures including VGG-verydeep16 and convolutional neural network 5 (CNN5) are evaluated for feature extraction and fusion. Experimental results show that hybrid model (VGG-verydeep16 + CNN5) achieves around 98% accuracy, outperforming individual models. The system adapts to dataset size, using CNN5 for large datasets and VGG-verydeep16 for smaller ones. This approach offers a robust, efficient, and scalable solution for real-world biometric authentication.
Volume: 15
Issue: 4
Page: 3252-3268
Publish at: 2026-08-01

Analyzing academic acceptance of artificial intelligence using extended technology acceptance model

10.11591/ijai.v15.i4.pp3090-3102
Nanang Suryadi , Abdurrahman Hakim , Adelia Shabrina Prameka , Wildan Syafitri , Muhammad Irfan Islami
This study aims to analyze the acceptance of artificial intelligence (AI) technology among Indonesian academics using extended technology acceptance model (TAM). The analysis involved general extended technology acceptance model for e-learning (GETAMEL) independent variables, which include subjective norm (SN), experience (EXP), enjoyment (ENJOY), computer anxiety (CA), and self-efficacy (SE), as well as mediator variables: perceived usefulness (PU) and perceived ease of use (PEOU). This study uses technology innovations (TI) as moderator variable and behavioral intention (BI) as a dependent variable. The analysis reveals that SN has a significant influence over PU but not PEOU. ENJOY and SE have a significant positive influence over both PU and PEOU, while EXP and CA don’t have a significant influence over both variables. PU and PEOU have a significant positive influence over BI. TI strengthens the relation between PU and BI, but weakens the relation between PEOU and BI. This finding provides an important outlook for developing a strategy to increase the acceptance of AI technology in the academic environment through an approach that takes into account the factors of pleasure, self-confidence, and TI.
Volume: 15
Issue: 4
Page: 3090-3102
Publish at: 2026-08-01

Long short-term memory based activity detection using skeleton joints data: a systematic review

10.11591/ijai.v15.i4.pp3026-3035
Bakkala Santha Kumar , R. Shankar
In today's security and surveillance applications, recognizing abnormal activity is critical component of identifying possible hazardous or unusual human behaviors. There is need for new technologies that can detect abnormal human behaviors precisely. The present review investigates various aspects of detection process, focusing on skeleton joints input data. It then explores adoption of deep learning (DL) architectures, such as long short-term memory (LSTMs) and transformers, to improve accuracy and robustness of recognition models. The review explores synergistic integration of LSTMs and transformers to improve recognition of unusual activity. By integrating LSTMs' processing capabilities and attention mechanisms of transformers, enhanced models can accurately identify intricate patterns of activity. Despite the advancements that have been made in the field, the challenges that remain are still related to recognition of unusual activities. These include lack of scalability for large datasets, need for models that can recognize complex behaviors across diverse applications, and need to ensure that detection is performed in a low-latency manner. The paper explores future directions of developing LSTM-based models that can recognize unusual activity using skeleton joint data in a cloud-based environment. The review emphasizes the potential of such solutions that can take advantage of the processing power of graphics processing units (GPUs) and tensor processing units (TPUs) and enable real-time recognition of activity in large datasets.
Volume: 15
Issue: 4
Page: 3026-3035
Publish at: 2026-08-01

A review of brain signal analysis in food perception studies

10.11591/ijai.v15.i4.pp3967-3980
Harish S. Sinai Velingkar , Roopa R. Kulkarni , Prashant Patavardhan , Abdul Haq Nalband
Food perception is more than just taste. It is a multisensory experience shaped by smell, texture, and even how the food looks. The human brain works to make sense of all these signals, influencing everything from cravings to food choices. This review looks at more than 40 studies that used electroencephalography (EEG) and other brainwave analysis tools to understand how to process food related stimuli. In order to identify patterns associated with attention, mood, and appetite, these studies looked at brain responses using techniques including event related potentials (ERPs) and spectral analysis. Current research shows that human desire for food, level of focus on it, and the feelings it evokes are all correlated with certain brain signals. However, majority of the studies investigated individual senses in isolation and focused on basic taste attributes such as sweet, salty, or bitter. The complexity of eating in real life, where several senses interact, is missed by this. Furthermore, a lot of research is lab based, short term, and does not represent real world situations. Future research must adopt practical, multimodal methods in order to fully understand how humans perceive food. Such studies could help improve food design, predict consumer behavior, and even support interventions for eating disorders.
Volume: 15
Issue: 4
Page: 3967-3980
Publish at: 2026-08-01

GWO-optimized sparse Bayesian least squares regression for direction-of-arrival estimation in MIMO networks

10.11591/ijai.v15.i4.pp3431-3440
Anne Gowda Aleri Byregowda , Babu Nallur Venkateshappa , Anughna Narayanaswamy
Accurate direction-of-arrival (DOA) estimation is a critical requirement for massive multiple-input multiple-output (MIMO) systems operating in fifth-generation (5G) and beyond (5G/B5G) wireless environments. Although sparse Bayesian learning (SBL)–based techniques have demonstrated improved robustness by exploiting signal sparsity, their performance is often limited by fixed hyperparameter selection, sensitivity to noise, and suboptimal residual error minimization. To address these challenges, this paper proposes an optimized sparse Bayesian least squares regression (SBLSR) framework in which grey wolf optimization (GWO) is employed to adaptively optimize Bayesian hyperparameters and regression coefficients. The proposed approach jointly enforces sparsity and minimizes estimation error, enabling robust DOA estimation under dynamic noise conditions and varying network density. Extensive simulations conducted in a massive MIMO environment demonstrate that the optimized SBLSR consistently outperforms conventional SBLSR and state-of-the-art benchmark techniques in terms of root mean square error (RMSE), closely approaching the Cramér–Rao lower bound (CRLB) across a wide range of signal-to-noise ratios, sensor configurations, and Monte Carlo trials. The findings validate that the suggested optimized SBLSR framework offers a noise-resilient solution for high-precision DOA estimation in practical massive MIMO and MIMO radar systems.
Volume: 15
Issue: 4
Page: 3431-3440
Publish at: 2026-08-01

An optimized deep learning framework for brain tumor classification using magnetic resonance imaging

10.11591/ijai.v15.i4.pp3722-3731
Komal Kumar Napa , Rajkumar Govindarajan , Senthil Murugan Janakiraman , Jayanthi Arumugam
Accurate and interpretable classification of brain tumors in magnetic resonance imaging (MRI) scans plays a crucial role in early diagnosis and effective treatment planning. This study introduces a deep learning (DL) framework based on a customized YOLOv5m architecture integrated with a bidirectional feature pyramid network (BiFPN) for multi-class brain tumor classification. The integration of BiFPN enhances multi-scale feature fusion, improving detection across varied tumor types, while YOLOv5m ensures real-time inference capabilities. To mitigate class imbalance, a class-weighted cross-entropy loss is adopted. The model is evaluated on multiple performance metrics, achieving a test accuracy of 88.86%, precision of 88.70%, recall of 88.20%, and F1-score of 88.25%. It also reports a mean average precision (mAP@0.5) of 94.36%, with high class-wise average precision (AP) for glioma, meningioma, pituitary, and no-tumor categories. Computational time for training (12484.81 seconds) and testing (146.82 seconds) confirms the model’s feasibility for real-time clinical deployment. To support interpretability, gradient-weighted class activation mapping (Grad-CAM) is integrated for visualizing class-discriminative regions, helping clinicians understand the model’s predictions. A gradio-based user interface is also developed, enabling intuitive interaction with the system.
Volume: 15
Issue: 4
Page: 3722-3731
Publish at: 2026-08-01

Emotion recognition of electroencephalogram using hybrid convolutional neural networks and vision transformer

10.11591/ijai.v15.i4.pp3934-3943
Esmeralda Contessa Djamal , Revan Vio Endriansyah , Daswara Djajasasmita
Electroencephalogram (EEG) signal-based emotion classification faces challenges in spatio-temporal complexity and inter-channel redundancy. This paper proposes a 2D convolutional neural network (CNN) and vision transformer (ViT) approach. Relevant emotions require signal extraction in the 4-40 Hz frequency band using the discrete wavelet transform (DWT). This study uses the SEED dataset from 15 subjects, with 62 channels reduced to 12. Every 5-second segment is decomposed using DWT into four frequency bands, which are then mapped to a 2D spatial representation for CNNs and ViTs. Experiment results show that the DWT-2D CNN-ViT achieves the best accuracy of 87% with a final loss of 0.087. In the meantime, the CNN-only model (85%), the ViT-only model (77%), and the DWT-only model (47%). Testing with several optimizers also shows that AdamW provides the highest performance with the fastest training time of 12.34 minutes. These results demonstrate that this integration can produce more efficient and stable learning. These findings indicate that the combination of DWT and the CNN-ViT hybrid architecture is effective for accurate, stable EEG-based emotion recognition and is potentially applicable to real-time emotion-monitoring systems.
Volume: 15
Issue: 4
Page: 3934-3943
Publish at: 2026-08-01

A machine learning framework for predicting and optimizing return on investment across marketing channels

10.11591/ijai.v15.i4.pp3528-3536
Chandra Chathura , Keerthan Saya , Sathishkumar Mani
In the current fast-paced competitive marketing environment, firms require data-centric methods to maximize their investments in several avenues. This research work applies machine learning techniques to estimate the return on investment (ROI) for marketing costs, which helps organizations in budgeting more effectively. Four models including random forest, extreme gradient boosting (XGBoost), gradient boosting, and linear regression were utilized for their accuracy in making predictions. Results showed that the highest accuracy was achieved by linear regression at 99.39%, random forest at 99.07%, gradient boosting at 99.01%, and XGBoost at 98.81%. It was further noted that digital marketing avenues such as social media and online stores gave the highest ROI, indicating that companies should prioritize digital marketing more than traditional marketing. On a practical level, this approach helps marketing team for choosing high performing channels since it estimates expected returns from each marketing channels and make smarter budget allocation. Yet, the study is done by using Kaggle dataset. In order to improve its generality, future research may use larger real-world datasets and extensive visualization techniques.
Volume: 15
Issue: 4
Page: 3528-3536
Publish at: 2026-08-01

Systematic review of fraud detection using AI and ML with an emphasis on telecommunication industry

10.11591/ijai.v15.i4.pp3269-3285
Soly Mathew , Sindi Rryta
The telecommunications industry is one of the top industries affected by fraudulent activities. Given the financial impact, on top of confidentiality breaches, security concerns, and reduced service quality as well as consumer dissatisfaction, there is an immediate need to implement effective fraud detection approaches. While there have been different fraud detection systems implemented, technological advancements as well as the improved techniques of fraudsters have made the traditional approaches no longer efficient. This paper aims to further investigate the use of artificial intelligence (AI) and machine learning (ML) to create efficient and advanced fraud detection models based on the strategy used, models applied, accuracy of the system, as well as future research work suggested. A systematic review of 50 papers was conducted. The most prevalent strategy was supervised one, majority of papers used software instead of hardware, and the most common ML models were artificial neural network (ANN), support vector machines (SVM), and decision tree.
Volume: 15
Issue: 4
Page: 3269-3285
Publish at: 2026-08-01

Prioritizing ransomware indicators of compromise using algorithmic scoring for enhanced threat detection

10.11591/ijai.v15.i4.pp3865-3877
Krishna Prasad D. Subramanya , Prasanna Kumar H. Ramakrishna
Ransomware is a very serious cybersecurity threat. Its attack methods are continually evolving, which means that it is not very easy to detect it quickly. A big challenge faced in the ransomware defense is how to prioritize indicators of compromise (IoC) efficiently. This paper introduces a hybrid IoC ranking scheme based on static and dynamic analysis of the behavior of commonly circulating ransomware variants. Each IoC is given a weighted score according to its significance for investigations based on actual patterns of occurrence and contextual behavior. Experimental results clearly separate high-confidence indicators from low-confidence ones. Deleting shadow copy and connecting to Tor2web receive the highest rank scores of about 99.99, while older behaviors like locking screen show lower relevance, around 73.11. The proposed algorithm has a linear time complexity of O(n·m) for score calculation and a bounded space complexity of O(n·m), allowing it to scale for large IoC sets. The findings show that this ranked IoC framework enhances early ransomware detection and helps prioritize responses based on evidence in current security systems.
Volume: 15
Issue: 4
Page: 3865-3877
Publish at: 2026-08-01

Symmetry index-based gait improvement prediction using a CNN-LSTM-attention framework

10.11591/ijai.v15.i4.pp3625-3636
Pushpalatha Obanna , Premkumar Ramesh
The shortcomings of existing clinical assessments are addressed by using a deep learning (DL) framework. This study will assess the gait improvement of lower limb fractured patients using the publicly available GaitRec dataset. The dataset contains the vertical ground reaction force (GRF) data used to compute biomechanical kinetic features. This framework considers hip, knee, ankle, and calcaneus fractures as lower limb fracture class, and the symmetry indices computation is done between affected and unaffected limbs, and their temporal changes were examined to analyze and track the rehabilitation progress of individual subject. The framework successfully identifies the reduction in the asymmetry between left and right leg features. The overall gait improvement in patients is computed using initial and final session composite asymmetry index (ASI) values and if it is at least 30% then overall gait is improved, demonstrating its robustness and clinical relevance. The suggested convolutional neural network (CNN)-long short-term memory (LSTM)-attention hybrid model outperformed with regression metrics of root mean square error (RMSE) of 1.09, mean absolute error (MAE) by 0.64, and R2 score of 0.98 and classification metrics of accuracy 97.1%, precision of 97.6%, recall of 96%, and F1-score of 96.8% by capturing both the local patterns and temporal dynamics of the gait data.
Volume: 15
Issue: 4
Page: 3625-3636
Publish at: 2026-08-01

Strategic optimization of artificial intelligence digital learning in informatics engineering

10.11591/ijai.v15.i4.pp2999-3008
Aan Ansori , Birru Muqdamien , Ahmad Tabrani , Reza Syafrizal , Sutanto Sutanto , Eko Wahyu Wibowo , Syifa Amara Dhestiyani
This study examines the strategic optimization of artificial intelligence (AI)-based digital learning in the Department of Informatics Engineering by analyzing the interaction between internal and external factors that influence successful implementation. Employing a qualitative approach based on strengths, weaknesses, opportunities, and threats (SWOT) analysis, this research utilizes the internal factor analysis summary (IFAS) and external factor analysis summary (EFAS) to assess institutional readiness, constraints, and strategic opportunities for AI integration in higher education. The results indicate that external factors exert a stronger strategic influence than internal factors, as reflected by a higher EFAS score (1.50) compared to the IFAS score (1.25). Key external opportunities include the increasing demand for AI competencies, supportive national policies, and opportunities for industry collaboration, while the main internal limitations are limited faculty expertise in AI and insufficient AI-specific learning resources. Based on these findings, this study formulates a SWOT-based strategic framework that converts analytical outcomes into practical recommendations for curriculum development, faculty capacity building, and the adoption of adaptive AI learning technologies. This research contributes a context-specific and empirically grounded strategic model that advances AI integration beyond descriptive analysis, supporting more effective, personalized, and sustainable digital learning, particularly within Informatics Engineering programs in Indonesian Islamic higher education institutions.
Volume: 15
Issue: 4
Page: 2999-3008
Publish at: 2026-08-01
1 2 3
...
2091 2092
Show 50 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