cover
Article Per Year (5 Year)
All
Home Page
OAI Link
Editorial Team
Contact
Reviewer
Google Scholar
Contact Name
Ramdan Satra
Contact Email
Ramdan Satra
Phone
-
Journal Mail Official
ramdan@umi.ac.id
Editorial Address
-
Location
Kota makassar,
Sulawesi selatan
INDONESIA
ILKOM Jurnal Ilmiah
Published by Universitas Muslim Indonesia
ISSN : 20871716     EISSN : 25487779     DOI : -
Core Subject : Science,
Computer Science & IT
ILKOM Jurnal Ilmiah is an Indonesian scientific journal published by the Department of Information Technology, Faculty of Computer Science, Universitas Muslim Indonesia. ILKOM Jurnal Ilmiah covers all aspects of the latest outstanding research and developments in the field of Computer science, including Artificial intelligence, Computer architecture and engineering, Computer performance analysis, Computer graphics and visualization, Computer security and cryptography, Computational science, Computer networks, Concurrent, parallel and distributed systems, Databases, Human-computer interaction, Embedded system, and Software engineering.
Arjuna Subject : -
Articles 633 Documents
 60   61   62   63   64 
Optimization of CNN Architectures through Fine-tuning for SIBI Classification Nur Hilmi Insan Muhammad; Abdul Syukur; Pujiono Pujiono
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.2830.378-392

Abstract

This research addresses the computational optimization of convolutional neural network (CNN) architectures for the classification of Indonesian Sign Language System (Sistem Isyarat Bahasa Indonesia, SIBI) static alphabet imagery to enhance digital communication accessibility. Utilizing a domain-specific dataset comprising 1,165 images across 26 alphabet classes, this study tackles the prominent challenges of limited sample sizes and severe class imbalance. We evaluate five state-of-the-art CNN architectures MobileNetV2, DenseNet121, Xception, InceptionV3, and ResNet50V2 under four distinct training data paradigms before and after adaptive fine-tuning. To eliminate predictive bias without pixel-level distortion, oversampling is operationalized via Latent Space SMOTE on flattened vector embeddings, combined with dynamistic runtime image augmentation. The experimental results reveal that MobileNetV2, when optimized through partial layer-freezing (locking 150 baseline layers) under the integrated augmentation and oversampling combination scenario, achieved the highest macro-classification accuracy of 98.30%. This architecture also demonstrated superior efficiency, reducing the computational training latency to 0.53 minutes. The findings underscore the strategic advantage of leveraging optimized lightweight networks like MobileNetV2 for domain-specific visual recognition tasks.
Lightweight Deep Learning Models for Lung Disease Classification Using Chest X-ray Images Abdullah Sholum; Aji Prasetya Wibawa; Ardhana Putra Agustavada; Dafa Fadhilah Hilmi; Felix Andika Dwiyanto
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.3351.302-319

Abstract

Lung disease classification from chest X-ray images using deep learning has attracted significant attention due to its potential to support rapid and automated medical diagnosis. However, many deep learning models require high computational resources, limiting their applicability in resource-constrained environments. This study evaluates the effectiveness of lightweight deep learning architectures for lung disease classification using chest X-ray images consisting of COVID-19, Normal, and Pneumonia classes. Three architectures were comparatively analyzed, including a baseline Convolutional Neural Network (CNN), EfficientNetV2, and MobileNetV2. All models were trained and evaluated under identical preprocessing and experimental conditions using a dataset of 697 chest X-ray images with a 60:20:20 split ratio for training, validation, and testing. Experimental results show that EfficientNetV2 and MobileNetV2 achieved classification accuracy up to 0.99 with AUC-ROC values approaching 1.0. In terms of computational efficiency, both lightweight architectures reduced trainable parameters by approximately 99.3% compared to the baseline CNN, decreasing from 576,851 to 3,843 trainable parameters through transfer learning with frozen pretrained layers. MobileNetV2 achieved this performance with 0.613 GFLOPs and a model size of 9,435 KB, while EfficientNetV2 required 0.781 GFLOPs and 15,993 KB. Although EfficientNetV2 demonstrated slightly more stable performance, MobileNetV2 provided the most effective trade-off between classification accuracy and computational efficiency, making it more suitable for deployment in resource-constrained healthcare environments. These findings demonstrate the feasibility of lightweight deep learning architectures for efficient and reliable lung disease classification from chest X-ray images
Design and Quantitative Evaluation of a Keycloak-Based Single Sign-On Architecture for Integrated Institutional Information Systems Nurhadi Nurhadi; Mustazzihim Suhaidi; Muhammad Athariq
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.3368.237-254

Abstract

The increasing integration of digital services in higher education institutions requires a secure and scalable authentication mechanism to ensure consistent access across multiple systems. However, fragmented authentication approaches often result in repeated login processes, inconsistent security policies, and inefficient identity management. This study proposes a modular Single Sign-On (SSO) architecture based on Keycloak, integrated with OAuth 2.0, OpenID Connect, and JSON Web Tokens (JWT), to support unified authentication in institutional information systems. A quantitative experimental approach is employed to evaluate system performance in a real academic environment involving 100 user accounts. The evaluation focuses on authentication efficiency, scalability, reliability, and user productivity. The results show a 62% reduction in average login time, an 85% increase in authentication throughput, and a 100% authentication success rate. Scalability testing indicates stable system performance under concurrent workloads, while token validation overhead remains minimal, ensuring that security enhancements do not degrade system responsiveness. In addition, task completion time decreases by 51%, accompanied by a significant improvement in user productivity. These findings demonstrate that the proposed Keycloak-based SSO architecture provides measurable improvements in performance, scalability, security governance, and usability. The study contributes to software systems engineering by presenting a validated architectural model and a comprehensive quantitative evaluation framework for identity management in higher education environments.
Comparative Analysis of Naïve Bayes Variants for Best-Seller Status Determination in the Air Conditioning Industry Sasa Ani Arnomo; Zada Alzena; Heri Nuryanto; Siti Fairuz Nurr Sadikan
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.3509.393-403

Abstract

The determination of product status such as the best seller predicate for Air Conditioning products is a fundamental sales promotion strategy. Leveraging machine learning to analyze sales data is essential for maximizing business progress and market positioning within the modern electronics industry. This study aims to evaluate and compare the performance and accuracy of Naïve Bayes (NB) classification models in determining AC product status. The goal is to identify the most effective variant among BernoulliNB, GaussianNB, and MultinomialNB for this specific application. A quantitative comparative analysis was conducted using various data record sizes. The primary features analyzed included brand, inverter type, electrical power cooling capacity, address, price, and sales status. In the first scenario, testing across varying data record sizes revealed that the MultinomialNB variant achieved the highest average accuracy at 85.39 percent. In the second scenario, using the full dataset with an 80% training data split without cross-validation, the Naïve Bayes algorithm as a whole demonstrated robust classification ability, reaching a peak prediction accuracy of 99.312 percent. This peak performance significantly outperformed alternative methods such as SVM and KNeighbors Classifier within those specified parameters. This performance significantly outperformed alternative methods such as SVM and KNeighbors Classifier within the specified parameters. The study concludes that the Naïve Bayes model is highly effective for product status classification in the electronics industry. The MultinomialNB variant is identified as the most consistently reliable model for these specific datasets. Future research should consider incorporating cross-validation techniques to further validate model stability across more diverse data environments.
Optimization of Intelligent Traffic Control Based on iot and Reinforcement Learning for Congestion Reduction in Smart Cities Tri Aristi Saputri; Budi Sutomo; Dimas Akbar Maulana; Hendika Purnomo
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.3260.320-332

Abstract

Traffic congestion has become a major challenge in Indonesian urban areas due to rapid vehicle growth and the limited adaptability of conventional traffic signal control systems. Most existing Deep Reinforcement Learning (DRL)-based traffic signal control studies adopt a free-phase selection approach, which assumes full agent freedom in determining signal phases — an assumption fundamentally incompatible with fixed phase-sequence regulations in Indonesian urban infrastructure — and rely on synthetic traffic data that fails to represent motorcycle-dominated traffic conditions. Furthermore, existing DQN-based approaches treat all traffic density conditions uniformly, without utilizing IoT-derived density categories for context-aware decision-making. To address these gaps, this study proposes a manual phase rotation mechanism with constrained actions (15, 30, and 60 seconds) compatible with existing fixed-phase infrastructure without hardware modifications, real-world IoT CCTV data from four intersections in Metro City processed using the YOLOv11 model to generate Low, Medium, and High traffic density categories as a representative training foundation for Indonesian urban traffic conditions, and a category-based action bias mechanism that adjusts DQN Q-value estimates according to IoT-derived traffic density, enabling context-aware signal duration selection. The DQN agent interacts with the SUMO simulation environment through the TraCI interface, receiving real-time traffic states comprising vehicle count, queue length, waiting time, average speed, density category, and delta queue, and selecting optimal green signal durations based on an epsilon-greedy exploration strategy and experience replay mechanism over 1,100 training episodes. Training yielded a 39.2% improvement in total reward and a 6.6% reduction in average waiting time. The best-performing model, obtained at episode 1050, achieved an 8.6% reduction in average waiting time and an 11.7% increase in traffic throughput compared to the fixed-time baseline. These results demonstrate that the proposed framework contributes three concrete advances for adaptive traffic signal control, a constrained-action DQN that is fully compatible with real-world fixed-phase infrastructure, a real-world IoT CCTV dataset as a representative data foundation for Indonesian traffic conditions, and a category-based bias mechanism for context-aware control — collectively offering a deployable, infrastructure-compatible, and replicable solution for traffic authorities and local governments advancing the smart city agenda in Indonesia.
Feature Importance–Driven Multimodal Learning for Medical Diagnosis Classification Using Clinical and Symptom Text Data Indra Waspada; Satriawan Rasyid Purnama; Alfonso Clement Sutantio; Alwey Hakim
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.3289.255-267

Abstract

The integration of structured clinical measurements and unstructured textual symptom descriptions poses persistent challenges for automated medical diagnosis, particularly due to feature heterogeneity and class imbalance in real-world outpatient data. This study proposes a feature importance–driven multimodal machine learning framework for multi-class medical diagnosis classification that jointly models numerical clinical attributes and free-text symptom narratives within a unified pipeline. Beyond overall performance comparison, the proposed approach systematically examines the interaction between model architecture and feature importance through three controlled configurations: base, strong-feature, and weak-feature settings. Six supervised learning algorithms are evaluated using stratified five-fold cross-validation and imbalance-aware metrics. The results show that feature importance–driven modeling yields strongly model-dependent performance characteristics. Bagging-based tree ensembles benefit most from strong-feature selection, with the Extra Trees classifier achieving the best overall performance, reaching a macro-averaged F1-score of 0.821, compared to 0.811 in the base configuration and 0.642 when only weak features are retained. Conversely, margin-based classifiers rely on distributed feature representations. The kernel-based Support Vector Machine performs poorly under strong-feature selection (F1-score 0.450) but achieves a substantially higher F1-score of 0.795 and a macro-averaged recall of 0.808 under the weak-feature configuration. Linear SVM demonstrates stable behavior across configurations, maintaining a macro-averaged F1-score between 0.798 and 0.810, while attaining the highest overall recall of 0.819. These findings indicate that feature importance should be treated as a model-aware analytical tool for aligning feature selection strategies with the inductive bias of the learning algorithm, supporting robust and clinically meaningful diagnostic classification.
Analysis of a Hybrid DNN–BiLSTM Framework for Longitudinal Prediction of Lung Disease Recurrence Using Clinical Data Olha Musa; Zainudin Sidik; Ifriandi Labolo; Muliati Badaruddin; Abdul Malik I. Buna
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.3222.404-419

Abstract

Predicting lung disease recurrence from longitudinal clinical data remains challenging because irregular temporal patterns and heterogeneous patient characteristics reduce the effectiveness of conventional deep learning models. This study analyzes a hybrid Deep Neural Network (DNN)–Bidirectional Long Short-Term Memory (BiLSTM) framework for the longitudinal prediction of lung disease recurrence using clinical data collected between 2021 and 2024 from a referral hospital in Gorontalo. The dataset includes demographic information, laboratory examination results, clinical diagnoses, and longitudinal medical records. Lung disease recurrence is defined as the reappearance or worsening of the disease during longitudinal clinical follow-up after the initial diagnosis or treatment. The proposed framework combines DNN to learn complex nonlinear relationships among multivariate clinical features and BiLSTM to capture temporal dependencies across sequential patient observations. Model performance was evaluated using Accuracy, Precision, Recall, F1-score, and Root Mean Square Error (RMSE), and compared with standalone DNN and BiLSTM models. Experimental results demonstrate that the proposed hybrid framework consistently outperformed the individual models, achieving an improvement of approximately 7–10% across the evaluation metrics while providing more stable longitudinal prediction performance. Furthermore, multivariate analysis identified dominant clinical variables associated with lung disease recurrence, improving the interpretability of prediction results for clinical decision-making. These findings indicate that the proposed framework provides an effective computational approach for longitudinal clinical prediction and supports the development of intelligent clinical decision support systems for recurrence risk assessment
IoT-Based Household Energy Monitoring with Preliminary Large Language Model-Assisted Recommendations Atikah Tri Budi Utami; Hamdan Gani; Annisa Dwi Damayanti
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.3233.333-350

Abstract

IoT-based household energy monitoring systems are effective for real-time data acquisition but often provide limited interpretive support for non-technical users. Most existing implementations focus on measurement and visualization, with less attention to converting raw consumption data into actionable recommendations. This study presents a prototype household energy monitoring system that integrates an ESP32-based IoT platform, PZEM-004T sensors, Firebase Realtime Database, and a cloud-hosted Large Language Model (LLM) in a hybrid edge-cloud architecture. Sensor validation was performed by comparing PZEM-004T readings with clamp meter measurements. The results showed low overall error, with voltage measurement achieving an MAE of 0.505 V, RMSE of 0.678 V, and MAPE of 0.231%, while power measurement achieved an MAE of 0.206 W, RMSE of 0.267 W, and MAPE of 0.889%. The LLM module was assessed through scenario-based functional evaluation, where it generated contextual explanations, identified potential energy-use anomalies, and produced quantified energy-saving suggestions. A real-user evaluation involving 50 respondents further indicated generally positive perceptions of the generated recommendations in terms of clarity, relevance, usefulness, trust, and behavioral intention, although the findings remain limited to perceived user responses rather than long-term behavioral outcomes. Overall, the results demonstrate the feasibility of employing an LLM as an interpretive layer in IoT-based household energy monitoring and indicate its potential to improve the accessibility of energy information for non-technical users.
Phishing Email Detection Using SVM with RBF Kernel Based on Manhattan Distance Asrianda Asrianda; Sujacka Retno; Beno Jange; Mansur Mansur
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.2806.268-280

Abstract

This study examines the performance of a Support Vector Machine (SVM) model with a Radial Basis Function (RBF) kernel for phishing email detection using Euclidean and Manhattan distance measures. The dataset consists of 3,600 email samples, including 2,400 legitimate emails and 1,200 phishing instances. The features are designed to capture both linguistic and structural characteristics of emails, including word count, vocabulary diversity, stop word usage, number of links and domains, presence of email addresses, spelling errors, and urgency-related terms. The experiments were conducted using two train-test split ratios, 80:20 and 70:30, combined with hyperparameter tuning of C and gamma across 15 iterations. The findings indicate that the Manhattan distance consistently outperforms the Euclidean distance, particularly in terms of recall and F1-score, which are critical for detecting the minority class. The model achieved a best accuracy of 78.33%, accompanied by noticeable improvements in recall and F1-score. These results suggest that the choice of distance function within the RBF kernel plays a crucial role in enhancing model sensitivity and generalization when dealing with imbalanced data. Furthermore, the iterative hyperparameter tuning process contributes significantly to improving both performance and model stability. Overall, the SVM-RBF approach with Manhattan distance provides an effective and reliable framework for phishing email detection in machine learning applications.
Optimization of Naive Bayes Algorithm Using Genetic Algorithm for Heart Failure Prediction Teguh Cahyono; Yogiek Indra Kurniawan; Bangun Wijayanto
ILKOM Jurnal Ilmiah Vol 18, No 2 (2026)
Publisher : Prodi Teknik Informatika FIK Universitas Muslim Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33096/ilkom.v18i2.3559.420-429

Abstract

Feature selection is often reported to improve clinical prediction, yet optimization and evaluation on the same data can produce optimistic estimates. This study provides a leakage-controlled and reproducible reassessment of genetic algorithm (GA) feature selection for Gaussian Naive Bayes prediction of heart disease. The open-Heart Failure Prediction dataset contains 918 observations, 11 predictors, and a binary target. Clinically implausible zero values in resting blood pressure and cholesterol were recoded as missing; imputation, scaling, one-hot encoding, GA selection, and model fitting were confined to training data. Performance was estimated using repeated nested stratified cross-validation with five outer folds repeated five times and four inner folds for GA fitness. The GA used an 11-bit chromosome, population size 12, 10 generations, 0.80 uniform-crossover probability, 0.08 bit-flip mutation probability, tournament selection of size three, and two elites. Across 25 held-out outer folds, baseline Naive Bayes achieved 84.66% accuracy (SD 2.01%), whereas GA-selected Naive Bayes achieved 84.31% (SD 2.40%). The paired mean difference was −0.35 percentage points, with a bootstrap 95% confidence interval of −0.72 to 0.05 percentage points and a Wilcoxon p-value of 0.120. AUC values were 0.912 and 0.907, respectively. The GA retained 8.44 of 11 predictors on average; ExerciseAngina, FastingBS, ST_Slope, and Sex were selected in every outer fold. These results do not support a material accuracy gain from GA selection, but demonstrate modest dimensionality reduction and reveal predictors that are stable under resampling. The study emphasizes nested evaluation, uncertainty reporting, and full parameter disclosure for credible optimization claims

Page 63 of 64 | Total Record : 633

 60   61   62   63   64 

Filter by Year

2016 2026


Reset
Filter By Issues
All Issue Vol 18, No 2 (2026) Vol 18, No 1 (2026) Vol 17, No 3 (2025) Vol 17, No 2 (2025) Vol 17, No 1 (2025) Vol 16, No 3 (2024) Vol 16, No 2 (2024) Vol 16, No 1 (2024) Vol 15, No 3 (2023) Vol 15, No 2 (2023) Vol 15, No 1 (2023) Vol 14, No 3 (2022) Vol 14, No 2 (2022) Vol 14, No 1 (2022) Vol 13, No 3 (2021) Vol 13, No 2 (2021) Vol 13, No 1 (2021) Vol 12, No 3 (2020) Vol 12, No 2 (2020) Vol 12, No 1 (2020) Vol 11, No 3 (2019) Vol 11, No 2 (2019) Vol 11, No 1 (2019) Vol 11, No 1 (2019) Vol 10, No 3 (2018) Vol 10, No 3 (2018) Vol 10, No 2 (2018) Vol 10, No 2 (2018) Vol 10, No 1 (2018) Vol 10, No 1 (2018) Vol 9, No 3 (2017) Vol 9, No 3 (2017) Vol 9, No 2 (2017) Vol 9, No 2 (2017) Vol 9, No 1 (2017) Vol 9, No 1 (2017) Vol 8, No 3 (2016) Vol 8, No 3 (2016) Vol 8, No 2 (2016) Vol 8, No 2 (2016) Vol 8, No 1 (2016) Vol 8, No 1 (2016) More Issue