cover
Article Per Year (5 Year)
All
Home Page
OAI Link
Editorial Team
Contact
Reviewer
Google Scholar
Contact Name
Dede Kurniadi
Contact Email
dede.kurniadi@itg.ac.id
Phone
+6287880007464
Journal Mail Official
jistics@aptika.org
Editorial Address
Green Garden Residence C-87, Kabupaten Garut, Provinsi Jawa Barat, Indonesia, 44151
Location
Kab. garut,
Jawa barat
INDONESIA
Journal of Intelligent Systems Technology and Informatics
Published by Aliansi Peneliti Informatika
ISSN : -     EISSN : 3109757X     DOI : https://doi.org/10.64878/jistics
Core Subject : Science, Engineering,
Computer Science & IT Control & Systems Engineering Engineering
The Journal of Intelligent Systems Technology and Informatics (JISTICS) is an international peer-reviewed open-access journal that publishes high-quality research in the fields of Artificial Intelligence, Intelligent Systems, Information Technology, Computer Science, and Informatics. JISTICS aims to foster global scientific exchange by providing a platform for researchers, practitioners, and academics to disseminate original findings, critical reviews, and innovative applications. The journal is published three times a year (March, July, November) and may also publish special issues on emerging topics.
Arjuna Subject : Ilmu Komputer (semua kategori) - Aplikasi Ilmu Komputer
Articles 23 Documents
 1   2   3 
Skin Tone Classification in Digital Images Using CNN For Make-Up and Color Recommendation Nurapipah, Nida; Yuliana, Siti Sarah
Journal of Intelligent Systems Technology and Informatics Vol 1 No 3 (2025): JISTICS, November 2025
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v1i3.29

Abstract

Human skin tone variation is an obstacle in the development of a digital beauty product recommendation system. The purpose of this study is to categorize skin tone into three groups (Black, Brown, and White). Using a Convolutional Neural Network (CNN) based on the refined EfficientNetB0 architecture on a balanced dataset of 1,500 facial images, each class consisting of 500 images. All images in the dataset have been resized to 224 × 224 pixels to match the model input and ensure data uniformity and compatibility with the EfficientNetB0 model architecture used. The dataset used was obtained from the Kaggle platform and processed through the normalization and augmentation stages. It was then evaluated through the validation process using the 5-fold cross-validation method. This model achieved a total accuracy level of 88.67%, with the white category demonstrating precision (0.93), recall (0.95), and F1-score (0.94), as well as the highest AUC of 0.99, indicating very satisfactory performance. Additionally, this system can offer personalized beauty product recommendations, including foundation shades, lipstick colors, and clothing color palettes, tailored to specific skin tones. This method enhances the user experience by providing accurate recommendations that adapt to various lighting conditions, making it suitable for use on digital beauty platforms.
Development of MyCare AI: A Dual-AI Mental Health Chatbot for Personalized Emotional Support Arief, Zaenal Syamsyul; Hamzah, Muhamad; Azham, Moch Nazham Ismul
Journal of Intelligent Systems Technology and Informatics Vol 1 No 2 (2025): JISTICS, July 2025
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v1i2.34

Abstract

Access to mental health services remains a critical challenge in Indonesia, primarily due to societal stigma and limited availability of professional support. In response to this issue, this study introduces MyCare AI. This web-based mental health chatbot platform combines a Bi-LSTM-based emotion classification model with a generative conversational model provided by Google Vertex AI. This Dual-AI architecture enables the system to detect user emotions from Indonesian text inputs and deliver real-time, contextually appropriate, and empathetic responses. The emotion classification model is trained on a balanced English-language dataset representing four key emotional states: sadness, suicidal ideation, fear, and anger. The system employs a translation mechanism to convert Indonesian input into English before classification and then uses the detected emotion to condition the response generation process dynamically. The model achieved a classification accuracy of 95%, outperforming comparable models based on BERT-SVM and conventional LSTM architecture. This platform is intended for individuals who require immediate, anonymous, and continuous emotional support, including users in underserved or remote communities. MyCare AI represents a scalable and practical solution for digital emotional assistance and lays the groundwork for future integration with professional mental health services and native-language support frameworks. Key limitations include the system's reliance on real-time translation and an English-based dataset, highlighting the need for future development of culturally specific models.
From Local Features to Global Context: Comparing CNN and Transformer for Sundanese Script Classification Agustiansyah, Yoga; Fauzi, Dhika Restu
Journal of Intelligent Systems Technology and Informatics Vol 1 No 2 (2025): JISTICS, July 2025
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v1i2.38

Abstract

The digital preservation of historical writing systems like Aksara Sunda is critical for cultural heritage, yet automated recognition is hindered by high character similarity and handwriting variability. This study systematically compares two dominant deep learning paradigms, Convolutional Neural Networks (CNNs) and Transformers, to evaluate the crucial trade-off between model accuracy and real-world robustness. Using a transfer learning approach, we trained five models (ResNet50, MobileNetV2, EfficientNetB0, ViT, and DeiT) on a balanced 30-class dataset of Sundanese script. Performance was assessed on a standard in-distribution test set and a challenging, independently collected Out-of-Distribution (OOD) dataset designed to simulate varied real-world conditions. The results reveal a significant performance inversion. While EfficientNetB0 achieved the highest accuracy of 96.9% on in-distribution data, its performance plummeted on the OOD set. Conversely, ResNet50, despite being lower in in-distribution accuracy, proved to be the most robust model, achieving the highest accuracy of 92.5% on the OOD data. This study concludes that for practical applications requiring reliable performance, the generalization capability demonstrated by ResNet50 is more valuable than the specialized accuracy of EfficientNetB0, offering a crucial insight for developing robust digital preservation tools for historical scripts.
Automatic Sentiment Annotation Using Grok AI for Opinion Mining in a University Learning Management System Julianto, Indri Tri; Sidqi, Muhammad Affan Al
Journal of Intelligent Systems Technology and Informatics Vol 1 No 3 (2025): JISTICS, November 2025
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v1i3.42

Abstract

Sentiment analysis has become an essential tool in evaluating user feedback on digital learning platforms. Understanding student sentiments toward Learning Management Systems (LMS) in higher education can offer critical insights for system development and service improvement. This study aims to evaluate the effectiveness of AI-assisted sentiment labeling using Grok AI and ChatGPT compared to manual labeling for sentiment classification of student opinions on LMS at Institut Teknologi Garut. The research involved distributing an online questionnaire to 96 students across four academic levels, collecting open-ended responses regarding their LMS usage experiences. These responses were preprocessed through case folding, cleaning, tokenization, stopword removal, and stemming. The sentiment labels were assigned using Grok AI, ChatGPT, and manual annotation, and the resulting datasets were used to build classification models using the Naïve Bayes algorithm in Altair RapidMiner with 10-Fold Cross Validation. The performance evaluation shows that manual labeling yielded the highest accuracy (52.22%) and Cohen's Kappa (0.137), followed by ChatGPT (50.11%, 0.119) and Grok AI (48.00%, 0.087). Word cloud visualizations further revealed the dominant themes within each sentiment class, indicating that positive opinions emphasized helpfulness and ease of use, while negative ones focused on access issues and system lags. This research suggests that AI-assisted labeling methods can be viable alternatives, although manual labeling still offers slightly better accuracy.
User Sentiment Analysis X Towards Makan Bergizi Gratis Program Using Automatic Labeling Technique with Deepseek AI Julianto, Indri Tri; Nurpajar, Dini Siti
Journal of Intelligent Systems Technology and Informatics Vol 1 No 2 (2025): JISTICS, July 2025
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v1i2.43

Abstract

Public perception of national nutrition initiatives is instrumental in shaping inclusive and data-driven policy development. In Indonesia, the "Makan Bergizi Gratis" (MBG) program introduced by President Prabowo has drawn significant attention, particularly on the X platform (formerly Twitter). This research topic was selected due to its national urgency and political significance, as the MBG program emerged as a key agenda during the 2024–2025 political transition. Therefore, examining public sentiment is essential to assess policy acceptance and identify areas for improvement. This study analyzes user sentiment toward the MBG policy using an automatic labeling approach supported by DeepSeek AI and the VADER Lexicon, followed by sentiment classification through the K-Nearest Neighbor (KNN) algorithm. The research involved five main stages: collecting 1,704 tweets from X between January 2024 and March 2025, preprocessing the text, conducting automatic sentiment labeling, applying TF-IDF for vectorization, handling class imbalance using the Synthetic Minority Over-sampling Technique (SMOTE), and classifying sentiments using KNN. The results indicate that without SMOTE, the VADER model achieved higher accuracy (93.49%) but lower Cohen's Kappa (0.16), while DeepSeek AI yielded lower accuracy (73.67%) but slightly higher Kappa (0.17). After SMOTE was applied, accuracy declined (VADER to 77.25%, DeepSeek AI to 64.72%), but Kappa scores improved significantly (VADER to 0.65, DeepSeek AI to 0.47), indicating more balanced and consistent sentiment predictions across classes. In conclusion, integrating automatic labeling, SMOTE, and KNN provides a reliable and scalable framework for analyzing large-scale sentiment on social media platforms, particularly in contexts with imbalanced opinion distributions.
Sentiment Analysis of the Residency Policy Launch in the New Student Admission System Using Automatic Labeling with Meta AI Julianto, Indri Tri; Lindawati
Journal of Intelligent Systems Technology and Informatics Vol 2 No 1 (2026): JISTICS, March 2026
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v2i1.44

Abstract

The launch of a domicile-based policy in Indonesia's New Student Admission System (SPMB) has triggered various public responses, especially on social media platforms. Understanding these sentiments is essential for evaluating policy acceptance and guiding future improvements in educational governance. This study aims to analyze public sentiment toward the policy using automatic labeling techniques and machine learning classification, with a focus on identifying dominant public perceptions. The research applies the CRISP-DM methodology, consisting of six stages: business understanding, data understanding, data preparation, modeling, evaluation, and deployment. A total of 1,105 comments were collected from Instagram and YouTube via web scraping and then preprocessed using text cleaning, stemming, and tokenization. Sentiment labels were generated using three automatic methods: Meta AI, RoBERTa, and TextBlob. Classification was performed using the Support Vector Machine (SVM) algorithm with four kernel variations. The results indicate that the combination of TextBlob labeling and an SVM with the Sigmoid kernel achieved the highest accuracy (0.99), along with strong precision, recall, and F1 Scores. Word cloud visualizations revealed that positive sentiment was related to educational access and teacher appreciation, while negative sentiment focused on dissatisfaction with fairness and system transparency. In conclusion, this study demonstrates that automated sentiment analysis, when supported by proper preprocessing and class balancing, is a powerful approach to extracting meaningful insights from public discourse. The findings are expected to support policymakers in developing data-driven strategies for improving future education policies.
A Deep Learning Method for Forest Fire Classification Using Convolutional Kolmogorov-Arnold Network Nur Sahid, Ahmad; Fauzi, Dhika Restu
Journal of Intelligent Systems Technology and Informatics Vol 2 No 1 (2026): JISTICS, March 2026
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v2i1.48

Abstract

Forest fires pose a significant threat, requiring advanced detection systems. Conventional deep learning models, such as Convolutional Neural Networks (CNNs), are often limited by fixed activation functions that struggle to model the complex, irregular visual patterns of fire. This architectural rigidity presents a research gap for more adaptive neural architectures. This study addresses this gap by proposing and evaluating a novel method for forest fire classification using a Convolutional Kolmogorov-Arnold Network (CKAN), an architecture featuring learnable activation functions to improve detection accuracy and flexibility. Following a systematic machine learning lifecycle, this research utilized a public Kaggle dataset of 14,063 'Fire' and 'Nofire' images. Extensive data augmentation was applied to enhance model robustness. We designed a hybrid CKAN model combining a CNN feature extractor with a KAN module that uses learnable B-spline activation functions for classification. The model was trained for 30 epochs with the AdamW optimizer and Binary Cross-Entropy loss, followed by a rigorous evaluation on an unseen test set. The proposed CKAN model demonstrated exceptional performance, achieving 98.04% accuracy and an AUC-ROC of 0.9955, significantly outperforming conventional architectures. Grad-CAM analysis confirmed that the model focused on relevant visual features of fire and smoke, thereby validating its decision-making process. The findings establish the CKAN architecture as a highly effective and computationally efficient approach for forest fire classification, making it a powerful and promising solution for deployment in real-world, resource-constrained environmental monitoring systems.
Load Testing-Based Performance Evaluation of the SiUKT API System Rachman, Andi Nur; Shofa, Rahmi Nur; Sjamsuddin, Irfan Nafis; Tarempa, Genta Najwar; Julianto, Indri Tri; Athoillah , Bifahmi Ahmad
Journal of Intelligent Systems Technology and Informatics Vol 2 No 1 (2026): JISTICS, March 2026
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v2i1.89

Abstract

Software testing is a crucial stage in the system development lifecycle. Previous studies on SiUKT have used only the black-box method, focusing on functionality without providing insights into performance optimization. This study aims to analyze and improve the performance of the SiUKT API using an intelligent load-testing approach with Apache JMeter. The testing measures three key indicators—response time, throughput, and error rate—across 10 API endpoints with concurrent user simulations of 10, 100, 250, and 500 users. The results show that the SiUKT website performs effectively under moderate load conditions, with an average response time of 338 ms and a throughput of 8.2 requests per second for 10 users. Under high load (500 users), performance declines, with response times ranging from 6 to 8 seconds, while throughput remains stable and the error rate stays at 0.00%. Only the register endpoint experienced a 100% error rate due to validation conflicts. These findings demonstrate the system's ability to maintain stability under varying loads and highlight performance degradation patterns as user traffic increases. The research contributes to the optimization of intelligent system performance by establishing quantitative benchmarks for API scalability and providing recommendations for adaptive infrastructure improvements to support automated intelligent load management.
Feature Importance Analysis for Predicting Residential Building Energy Consumption Using Random Forest Tiara; Rahayu, Sri Nur
Journal of Intelligent Systems Technology and Informatics Vol 2 No 1 (2026): JISTICS, March 2026
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v2i1.99

Abstract

Using the Random Forest algorithm within the CRISP-DM framework, this study investigates residential building energy consumption and identifies key factors influencing electricity demand. The dataset, obtained from Kaggle, includes environmental, operational, and temporal variables, such as temperature, humidity, occupancy, HVAC usage, lighting activity, renewable energy utilization, and hour_of_day. Data preprocessing and categorical encoding were applied prior to model training, followed by hyperparameter optimization using RandomizedSearchCV. The optimized Random Forest model achieved a Mean Absolute Error (MAE) of 4.349, a Root Mean Square Error (RMSE) of 5.499, and a Coefficient of Determination (R²) of 0.510, indicating moderate predictive performance. While the model effectively captures general consumption patterns, part of the variability remains unexplained due to the absence of explicit temporal dependencies and detailed behavioral factors. Model interpretability was examined using Feature Importance, SHAP, and Partial Dependence Plots (PDP), revealing that temperature and HVAC usage are the most influential predictors of residential energy consumption. Overall, the proposed approach provides interpretable insights into residential energy use patterns and supports data-driven strategies to improve energy efficiency in residential buildings.
Comparison of Score-Based and Content-Based Automatic Sentiment Labeling Using a K-Nearest Neighbor Classifier Muzaky, Rifky Khoerul; Diniyaturobiah, Hanipah
Journal of Intelligent Systems Technology and Informatics Vol 2 No 1 (2026): JISTICS, March 2026
Publisher : Aliansi Peneliti Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64878/jistics.v2i1.120

Abstract

This study investigates the performance gap between two automatic sentiment labeling strategies one relying on star ratings and the other derived from textual content in classifying application reviews using the K-Nearest Neighbor (KNN) algorithm. Each review is converted into TF-IDF vectors, and the influence of both labeling approaches on the resulting classifier is examined. Performance is evaluated using accuracy, precision, recall, and F1-score to ensure a comprehensive assessment, with the content-based method achieving an accuracy of 0.81, indicating a more reliable outcome than the score-based variant. The score-driven approach shows weaker consistency, largely due to mismatches between numerical ratings and the sentiment conveyed in written text. Despite these findings, the study is limited by its focus on a single application domain and its reliance on a single classical baseline classifier, which may be sensitive to class imbalance. Future work is encouraged to incorporate more diverse datasets, adopt modern text representation techniques such as word embeddings or transformer-based encodings, and explore classification algorithms that better accommodate uneven class distributions.

Page 2 of 3 | Total Record : 23

 1   2   3 

Filter by Year

2025 2026


Reset
Filter By Issues
All Issue Vol 2 No 1 (2026): JISTICS, March 2026 Vol 1 No 3 (2025): JISTICS, November 2025 Vol 1 No 2 (2025): JISTICS, July 2025 Vol 1 No 1 (2025): JISTICS, March 2025