Articles

Forecasting income inequality in Vietnam’s regions using machine learning

10.11591/ijai.v15.i4.pp3464-3475

This study applies artificial neural networks (ANN) and support vector regression (SVR) to forecast the Gini coefficient across Vietnam’s seven socio-economic regions using limited data from 2014–2022 (N = 63). Despite the small dataset, robust techniques including repeated cross-validation, temporal backtesting, regularization, and bootstrapping were employed to mitigate overfitting. The results show a modest continued decline in income inequality through 2025, with faster improvements in urban regions (Red River Delta and Southeast) than in rural and highland areas. ANN outperformed SVR with an average cross-validation coefficient of determination (R²) of 0.93. The novelty lies in its granular regional forecasting and region-specific policy recommendations beyond national-level analyses. Key drivers—per capita income and simple housing—support targeted interventions such as housing credits in the Mekong River Delta and agricultural investments in the Central Highlands. The findings provide an evidence-based foundation for regionally tailored strategies to reduce income disparities in Vietnam.

Volume: 15

Issue: 4

Page: 3464-3475

Publish at: 2026-08-01

Hybrid deep learning model for the task scheduling in cloud computing

10.11591/ijai.v15.i4.pp3683-3691

Kavita Rani , Om Prakash Sangwan , Ritu Garg

Task scheduling plays a crucial role in optimizing performance, reducing costs, and enhancing system reliability by efficiently allocating resources to workloads. Traditional task scheduling methods lack the ability to efficiently manage workloads and resource distribution, leading to potential inefficiencies in performance and energy consumption. To address these limitations, advanced techniques leveraging deep learning and reinforcement learning are explored. This study proposes a deep learning-based model for task scheduling in cloud computing. The model employs a convolutional neural network (CNN) to predict the optimal machines for task allocation. Additionally, Q-learning is integrated with CNN to facilitate load shifting between machines, ensuring efficient utilization of resources. The dataset used in this work consists of task attributes, such as execution time, resource requirements, which were loaded from a CSV file. Comparative analysis with existing models shows that the proposed approach achieves approximately 94% accuracy and consumes less energy than other models, demonstrating its effectiveness in cloud task scheduling.

Volume: 15

Issue: 4

Page: 3683-3691

Publish at: 2026-08-01

Immuno-bioinformatics analysis of progressive alignment and logic learning machine-derived viral conserved sequences

10.11591/ijai.v15.i4.pp3189-3201

Felza Ridho , Mohammad Isa Irawan , Nurul Hidayat , Awik Puji Dyah Nurhayati

This study proposes a method to identify conserved sequence positions in viral data, with the primary goal of facilitating their translation into proteins. The approach aims to support the early detection of sub-sequences that hold promise as vaccine candidates. The method involves five key steps. First, mutation analysis using the Kimura model: analyzed mutations in the viral data using the Kimura model to provide insights into sequence variations. Second, alignment method selection to align sequences effectively, the progressive alignment approach with the neighbor-joining algorithm was chosen. Third, hybrid algorithm for identifying unchanged sequences: a hybrid algorithm that combines progressive alignment and a logic learning machine (LLM) was employed to identify unchanged sequences. Fourth, determining conserved sequences: based on the longest unchanged sequence, conserved regions within the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) data were identified. Fifth, immuno-bioinformatics analysis for vaccine candidate peptides: using immuno-bioinformatics, protein sequences were analyzed to identify potential vaccine candidates based on B cell epitopes. Experimental validation confirmed the algorithm’s ability to identify candidate proteins and generate vaccine-worthy peptides from conserved protein sequences. The streamlined approach focuses on conserved protein sequences, ensuring efficient and targeted identification of vaccine candidates. By leveraging these conserved regions, contributions can be made to the expedited development of vaccines.

Volume: 15

Issue: 4

Page: 3189-3201

Publish at: 2026-08-01

Analysis of normalization technique on multi objective preference analysis method

10.11591/ijai.v15.i4.pp3827-3835

Fristi Riandari , Gabriel Ardi Hutagalung , Ferry Fachrizal

Normalization is a critical step in multi-criteria decision analysis (MCDA) because it influences ranking consistency and decision reliability. This study evaluates the effects of four normalization techniques linear max, linear max-min, linear sum, and semi-linear vector, within the multi objective preference analysis (MOPA) framework using a tourism development case involving 18 alternatives and 8 decision criteria with both cost and benefit attributes. The techniques were compared based on ranking behavior, discriminative capability, and robustness using statistical and non-parametric validation. The results show that linear max-min normalization provides the strongest discriminative performance and the most significant statistical results, while semi-linear vector demonstrates high ranking stability and balanced sensitivity. In contrast, linear sum and linear max exhibit lower discriminative capability under the evaluated conditions. Kendall's tau and robustness analyses further confirm that normalization choice significantly affects ranking consistency and decision reliability. These findings provide practical guidance for selecting appropriate normalization techniques and support the development of more reliable MCDA-based decision-making models for complex applications, including sustainable tourism planning.

Volume: 15

Issue: 4

Page: 3827-3835

Publish at: 2026-08-01

Intravenous immunoglobulin resistance prediction in Kawasaki disease using oversampled transformer embeddings

10.11591/ijai.v15.i4.pp3944-3954

Namitha Thattarassery Nanappan , Raghavendra Srinivasaiah , Vinith Rejathalal

Kawasaki disease (KD) is a leading cause of acquired heart disease in children under five. Although intravenous immunoglobulin (IVIG) treatment is usually effective, 10–20% of cases are resistant and at higher risk for coronary complications. Early prediction of IVIG resistance is critical but difficult due to the rarity of KD and imbalanced clinical data. To address this, we propose a novel technique called sentence transformer embeddings with synthetic minority over-sampling technique (SMOTE) oversampling (STESO), which leverages the complementary strengths of transformer-based representation learning and synthetic oversampling. Pretrained models such as paraphrase-MiniLM-L3-v2 are used to convert tabular clinical data into dense text-based embeddings, capturing deeper semantic relationships across features. By coupling these rich embeddings with SMOTE, we balance class distributions directly in the semantic space, enabling traditional machine learning (ML) models to more effectively detect minority (resistant) cases. This synergy yielded substantial improvements in sensitivity and F1-score, with random forest (RF) combined with STESO (RF-STESO) achieving the highest overall performance. Among the models evaluated, our proposed model attained best result as accuracy of 0.85, a sensitivity of 0.81, a specificity of 0.89, and an F1-score of 0.85. Our results underscore that the joint use of transformer embeddings and oversampling is more effective than either approach in isolation, offering a promising pathway for rare disease prediction tasks such as IVIG resistance prediction in KD.

Volume: 15

Issue: 4

Page: 3944-3954

Publish at: 2026-08-01

Enhancing pedestrian detection in adverse conditions using YOLOv5s with adaptive weighted fusion efficient channel attention module

10.11591/ijai.v15.i4.pp3299-3308

Oumayma Rachidi , Badr Bououlid Idrissi , Chafik Ed-dahmani

Pedestrian detection constitutes a critical task within advanced driver assistance systems (ADAS), where reliable identification of pedestrians is essential for ensuring vehicular safety. Although deep learning has substantially improved detection performance, existing state-of-the-art models continue to exhibit notable degradation in adverse weather conditions and low-light. To mitigate these challenges, this study introduces an enhanced pedestrian detection framework based on you only look one version 5 (YOLOv5s), retrained on an augmented common object in context (COCO) dataset focused on the person class. Additionally, a novel, lightweight, and adaptive attention mechanism called: the weighted fusion efficient channel attention (WF-ECA) module is incorporated into the detection architecture. The WF-ECA module selectively focusses on important features without compromising computational efficiency or inference speed. Comparative experiments demonstrate a 5% increase in mean average precision (mAP) in comparison to the baseline model, thereby demonstrating the efficacy of the proposed attention module in improving detection robustness under challenging environmental conditions. These findings highlight the potential of attention-based mechanisms to enhance pedestrian detection performance in real-world ADAS applications.

Volume: 15

Issue: 4

Page: 3299-3308

Publish at: 2026-08-01

An empirical comparison of clustering approaches for recency, frequency, and monetary customer segmentation

10.11591/ijai.v15.i4.pp3402-3410

Upekkha Lau , Meditya Wasesa

This study evaluates effectiveness of three clustering techniques—k-means, hierarchical clustering, and density-based spatial clustering of applications with noise (DBSCAN)—applied to the recency-frequency-monetary (RFM) model for customer segmentation in the retail sector. Using sales transaction data from a distributor of computer accessories and printing products. The results show that k-means achieved the best clustering validation scores and effectively identified high-value customers, hierarchical clustering generated less meaningful groupings than k-means, and DBSCAN misclassified key customers as noise. These findings highlight k-means as the most suitable technique for RFM-based segmentation in this retail business context. The study offers practical insights for retail and distribution businesses aiming to adopt data-driven customer strategies and suggests future research to enhance segmentation robustness and refine the RFM framework.

Volume: 15

Issue: 4

Page: 3402-3410

Publish at: 2026-08-01

A transfer learning-based approach for automatic monument detection

10.11591/ijai.v15.i4.pp3326-3341

Santosh Giri , Jebish Purbey , Sunil Adhikari , Paarit Pokharel , Babu R. Dawadi , Bipun Man Pati , Atsushi Ito , Sushant Chalise , Sanjivan Satyal

Architectural heritage connects us to the cultural achievements of past civilizations. Patan Durbar Square in Nepal is home to many such structures, yet identifying them remains a challenge for tourists. This paper presents an automated monument recognition system built on the backbone of convolutional neural networks (CNNs). A dataset of 1832 images of 9 important monuments from Patan was created, and build a detection system with MobileNetV2, a light-weight CNN, to detect monuments in Patan Durbar Square with a near-perfect F1 score of 98.94%. The approach utilizes transfer learning to adapt the model to local architectural styles. A detailed ablation study is performed to determine the optimal network design and augmentation strategies. Class-wise performance is further analyzed to verify robustness against visual occlusion and similarity. Finally, the model is deployed as a mobile application using Flutter and the FastAPI framework. This work demonstrates the viability of lightweight CNNs for real-time cultural heritage preservation.

Volume: 15

Issue: 4

Page: 3326-3341

Publish at: 2026-08-01

Remaining useful life estimation for predictive battery maintenance with improved recurrent singular spectrum analysis algorithm

10.11591/ijece.v16i4.pp1817-1831

Chutipongse Boonyakitmaitree , Suchada Sitjongsataporn

As the global electric vehicle (EV) battery market is projected to reach a valuation of over USD 100 billion by the end of 2026, the demand for sophisticated battery management systems (BMS) has become more critical than ever. Accurate remaining useful life (RUL) prediction is essential for ensuring vehicle safety, optimizing maintenance, and evaluating retired batteries for second-life applications. However, existing prognostic methods often struggle to balance computational efficiency with predictive accuracy, especially during the early stages of battery usage. This research proposes combined weighted similarity-based and recurrent singular spectrum analysis (CWS-RSSA), a hybrid forecasting framework that integrates RSSA with a similarity-based approach through a weighted logistic switching mechanism. The algorithm is designed to be computationally lightweight, making it suitable for resource-constrained BMS hardware. The proposed method was validated using NASA and a large-scale dataset from MIT-Stanford consisting of 124 lithium-ion cells. Experimental results demonstrate that CWS-RSSA is capable of early-stage prediction with a relative error of 19.8%, whereas existing methods are unable to provide predictions. In later stages, once sufficient data becomes available, the algorithm achieves near-perfect accuracy with a negligible relative error on the NASA dataset and an average relative error of only 0.14% across the 124 MIT-Stanford batteries. Furthermore, the algorithm demonstrates robust performance in handling capacity regeneration phenomena. These findings suggest that CWS-RSSA represents a scalable and practical advancement for battery health management, supporting the transition toward a sustainable circular energy economy and providing a reliable foundation for second-life battery certification.

Volume: 16

Issue: 4

Page: 1817-1831

Publish at: 2026-08-01

Adaptive binary capsule network for complex image recognition

10.11591/ijai.v15.i4.pp3745-3760

Mavis Serwaa Yeboah , Patrick Kwabena Mensah , Adebayo Felix Adekoya , Mighty Abra Ayidzoe

Glaucoma and cataracts are leading causes of blindness worldwide, emphasizing the need for early detection. Convolutional neural networks (CNNs) have shown promise in detecting ocular diseases, but require extensive training datasets. However, medical datasets are scarce, limited, and imbalanced, prompting the use of time-consuming data augmentation techniques. To address these limitations, a computationally efficient and robust capsule network (CapsNet) model was proposed. The model features a novel adaptive contrast spatial filtering (ACSF) algorithm and incorporates a local binary pattern (LBP) algorithm to enhance robustness. This study achieved good recognition accuracy, with scores of 96.90% on the combined dataset, 96.45% on the cataract-only dataset, and 96.78% on the glaucoma-only dataset. The model's performance is comparable to state-of-the-art models, demonstrating its potential to support ophthalmologists in diagnosing cataracts and glaucoma-related eye issues.

Volume: 15

Issue: 4

Page: 3745-3760

Publish at: 2026-08-01

A bibliometric analysis of artificial intelligence applications in geospatial and geographic intelligence

10.11591/ijai.v15.i4.pp3036-3052

Chatarina Muryani , Singgih Prihadi , Gentur Adi Tjahjono , Albertus Erico Jerry Krisna Nugroho , Fitria Dewi Kartika

This study presents a comprehensive bibliometric analysis of the application of artificial intelligence in geospatial research and geographic intelligence. Using the Scopus database and visualization tools such as VOSviewer and Microsoft Excel, this study systematically maps the scientific literature across the five analytical dimensions: publication trends, citation structure, co-citation patterns, bibliographic merging, and keyword co-occurrence. The findings indicate a significant increase in global research interest, especially after 2020, with major contributions from the United States, China, and the United Kingdom. The results also reveal thematic clusters ranging from remote sensing and disaster response to smart city planning and spatial prediction. Through science mapping and performance analysis, the study highlights the intellectual structure and conceptual evolution of the field. The study contributes to academic understanding by identifying research gaps, emerging themes, and collaboration patterns, which can guide future interdisciplinary research in the artificial intelligence and geospatial domain.

Volume: 15

Issue: 4

Page: 3036-3052

Publish at: 2026-08-01

SHAP-enhanced ensemble learning for yield prediction: insights from CY-Bench data on Indian wheat

10.11591/ijai.v15.i4.pp3411-3420

Soma Gupta , Dayal Kumar Behera , Satarupa Mohanty , Subhra Swetanisha , Ritik Mallik , Namita Panda

Predicting crop yield accurately is essential for providing food security and improving agricultural practices. This study examines the use of ensemble machine learning models combined with Shapley additive explanations (SHAP) feature selection to improve wheat yield prediction in India. The study uses CY-Bench data, incorporating normalized difference vegetation index (NDVI), meteorological data, and soil moisture data for yield prediction. Various ensemble techniques, including voting, stacking, and boosting are evaluated. Boost m1 ensemble model consistently outperforms other models in the prediction. Additionally, the integration of SHAP-based feature selection with the best ensemble model significantly improves the model accuracy and interpretability by identifying the most influential features affecting yield. The results show the effectiveness of ensemble boosting model, in capturing the complex relationships within agricultural data particularly when combined with feature selection. This method improves the transparency and actionability of machine learning models for agronomists, policymakers, and farmers.

Volume: 15

Issue: 4

Page: 3411-3420

Publish at: 2026-08-01

EEB7-UNet: a deep learning framework for automated segmentation of fractured C-spine vertebrae

10.11591/ijai.v15.i4.pp3770-3781

Abhishek Kumar Pandey , Pateel G. P. , Kedarnath Senapati

Accurate identification of vertebral fracture (VF) regions in computed tomography (CT) images is crucial for surgeons prior to treatment planning, but remains challenging due to irregular vertebral boundaries, low contrast, noise, and image unevenness. Recent advancements in deep learning have shown promising results compared to conventional manual diagnosis methods in detecting anomalies and segmenting regions of interest in medical imaging. In this study, a deep learning model, enhanced EfficientNetB7 U-Net (EEB7-UNet), is proposed to segment the fractured cervical vertebrae. It includes custom data augmentation to increase the data size and a hybrid learning rate scheduler strategy technique for faster convergence, which increases the generalizability and robustness of the model. The proposed model achieved an improved dice score index of 95.53% and Jaccard coefficient index of 93.85% on the test dataset. Furthermore, the EEB7-UNet has emerged as a moderate size with 98.6 MB. The approach yields superior performance in terms of dice score index and Jaccard coefficient index compared to the other state of the art convolutional neural network (CNN) used as an encoder in the U-Net. This research also compared the performance of the proposed model with three other studies in similar contexts, reported in the literature.

Volume: 15

Issue: 4

Page: 3770-3781

Publish at: 2026-08-01

Detection of illegal drug activities in Indonesia via social media X

10.11591/ijai.v15.i4.pp3376-3388

Adi Hanif Sedar , Lukman Yudokusumo , Indra Budi , Amanah Ramadiah , Aris Budi Santoso , Prabu Kresna Putra

In today's digital world, social media platforms such as X (formerly Twitter) are widely used for communication but also for illegal activities, such as selling illicit drugs. This study examines how posts from X related to illegal drug sales in Indonesia can be analyzed using machine learning techniques, with a focus on incorporating features related to personally identifiable information (PII) and web URLs. The main questions are how to use classification methods to detect these posts and which drugs are most often mentioned. A total of 11,777 posts from X were collected using web scraping. After cleaning and processing the data, five machine learning models were used. The support vector machine (SVM) model showed the best results with a weighted F1-score of 97.08%. The findings show that abortion pills are the most frequently mentioned drugs, followed by sexual enhancement and anesthetic drugs. This study also used visual tools to show the most common hashtags, posts distribution over time, and types of drugs discussed. This study underlines the importance of monitoring social media to protect public health and highlights how incorporating PII features can improve the detection of illegal activities. Future research should look at posts in other languages and from other platforms, considering PII features for enhanced analysis.

Volume: 15

Issue: 4

Page: 3376-3388

Publish at: 2026-08-01

Deep learning-based prognostic modeling of brain tumors

10.11591/ijai.v15.i4.pp3792-3804

Laiali Almazaydeh , Arar Al Tawil

Headline accuracy is not enough to move a deep-learning brain tumor magnetic resonance imaging (MRI) classifier into the clinic. A model also needs repeated-split validation, trustworthy probabilities, a statistical sanity check, and a small enough footprint for hospital hardware. This study cover all four in one framework. Five convolutional neural network (CNN) backbones (visual geometry group (VGG)-16, residual network (ResNet)-50V2, MobileNetV2, EfficientNetB0, and dense convolutional network (DenseNet)-121) are trained on 4,600 public brain MRI scans (2,513 tumor, 2,087 healthy) under a shared two-phase transfer-learning recipe. The usual single split is replaced by stratified 5-fold cross-validation (CV), with paired McNemar and DeLong tests and 1,000 bootstrap resamples. Calibration is judged by expected calibration error (ECE), Brier, reliability diagrams, and temperature scaling. VGG16 wins mean accuracy (98.72 ± 0.42%); ResNet50V2 has the tightest area under the curve (AUC) (0.9986 ± 0.0006); the two are statistically tied. MobileNetV2 is the best-calibrated model out of the box (ECE = 0.0021) with only 2.59 M parameters, making it the most deployable. A four-level confidence-based risk score turns calibrated probabilities into triage tags. This study also flag a measured 3.23% format-duplicate leakage in the public dataset and presents the framework as a radiologist co-pilot, not a prognostic model.

Volume: 15

Issue: 4

Page: 3792-3804

Publish at: 2026-08-01

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

Filters article

Years

Source Title

Authors

31,458 Article Results

Forecasting income inequality in Vietnam’s regions using machine learning

Hybrid deep learning model for the task scheduling in cloud computing

Immuno-bioinformatics analysis of progressive alignment and logic learning machine-derived viral conserved sequences

Analysis of normalization technique on multi objective preference analysis method

Intravenous immunoglobulin resistance prediction in Kawasaki disease using oversampled transformer embeddings

Enhancing pedestrian detection in adverse conditions using YOLOv5s with adaptive weighted fusion efficient channel attention module

An empirical comparison of clustering approaches for recency, frequency, and monetary customer segmentation

A transfer learning-based approach for automatic monument detection

Remaining useful life estimation for predictive battery maintenance with improved recurrent singular spectrum analysis algorithm

Adaptive binary capsule network for complex image recognition

A bibliometric analysis of artificial intelligence applications in geospatial and geographic intelligence

SHAP-enhanced ensemble learning for yield prediction: insights from CY-Bench data on Indian wheat

EEB7-UNet: a deep learning framework for automated segmentation of fractured C-spine vertebrae

Detection of illegal drug activities in Indonesia via social media X

Deep learning-based prognostic modeling of brain tumors

Discover Our Library