cover
Article Per Year (5 Year)
All
Home Page
OAI Link
Editorial Team
Contact
Reviewer
Google Scholar
Contact Name
Hanis Amalia Saputri
Contact Email
editor.ijcshai@binus.edu
Phone
+6221-5345830
Journal Mail Official
editor.ijcshai@binus.edu
Editorial Address
https://journal.binus.ac.id/index.php/ijcshai/about/editorialTeam
Location
Kota adm. jakarta barat,
Dki jakarta
INDONESIA
International Journal of Computer Science and Humanitarian AI
Published by Universitas Bina Nusantara
ISSN : 30644372     EISSN : -     DOI : https://doi.org/10.21512/ijcshai.v2i2.14418
Core Subject : Humanities, Science, Engineering,
Humanities Computer Science & IT Control & Systems Engineering Electrical & Electronics Engineering Engineering
International Journal of Computer Science and Humanitarian AI (IJCSHAI) is an international journal published biannually in February and October. The Journal focuses on various issues: Computer Science, Artificial Intelligence (AI), Fuzzy Systems, Expert Systems, Geo-AI, Machine Learning, Deep Learning, Humanitarian AI, Data Science, Computer Vision, Natural Language Processing (NLP), Information Systems, Psychoinformatics, Computational Intelligence, Recommender Systems, Robotics, Robot Vision and Control Systems
Arjuna Subject : Ilmu Komputer (semua kategori) - Kecerdasan Buatan
Articles 27 Documents
 1   2   3 
Comparative Analysis of Decision Tree, Random Forest, and XGBoost for Student Category Prediction Lim, Rayson Calvianto; Harvianto, Harvianto
International Journal of Computer Science and Humanitarian AI Vol. 3 No. 1 (2026): IJCSHAI (In Press)
Publisher : Bina Nusantara University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21512/ijcshai.v3i1.14936

Abstract

This research aims to develop and evaluate a lightweight machine learning framework for predicting student performance categories as a foundation for personalized curriculum design in a mid-sized school context. The study compares three baseline algorithms such as Decision Tree, Random Forest, and XGBoost implemented using an end-to-end workflow involving data preprocessing, feature engineering, model training, and evaluation. A dataset of anonymized student academic and behavioral attributes was prepared through cleaning, encoding, normalization, and stratified splitting to ensure consistency and reliability. Each model was assessed using accuracy, precision, recall, and F1-score to determine its predictive effectiveness. The experimental results show that the Random Forest model achieved the highest overall performance, demonstrating stronger generalization compared to Decision Tree and XGBoost. Medium-performing students were classified most reliably, while Low-performing students displayed greater variability, indicating the need for more comprehensive data to improve sensitivity toward at-risk learners. The originality of this study lies in its focus on implementing an accessible, resource-efficient predictive pipeline suitable for schools with limited technological capacity. The findings provide evidence that practical machine learning approaches can support early stages of data-driven curriculum planning and help educators make more informed instructional decisions. The study also highlights opportunities for future work, including the expansion of data sources and adoption of more advanced algorithms to enhance predictive accuracy and support broader educational applications.
Optimizing CNN-LSTM Models for Stock Price Prediction in a Multi-Sector Holding Company Suprapto, Muhamad Fajar; Andryana, Septi
International Journal of Computer Science and Humanitarian AI Vol. 3 No. 1 (2026): IJCSHAI (In Press)
Publisher : Bina Nusantara University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21512/ijcshai.v3i1.15060

Abstract

Accurate stock price forecasting is difficult because financial time-series data usually demonstrate nonlinear relationships, irregular fluctuations, and interdependent temporal patterns. This research investigates the predictive performance of three neural network models based on deep learning: Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and a hybrid CNN-LSTM architecture for forecasting stock prices of a multi-sector holding company. The dataset used in this study contains daily historical price observations collected from 2015 to 2025, where sequential samples are generated using a sliding window approach. To obtain appropriate model settings, hyperparameter optimization is carried out using a grid search procedure. Model performance is assessed using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). Experiments are first performed using an 80:20 training-testing split and followed by a robustness evaluation using a 70:30 data split. Under the primary evaluation scheme, the experimental results indicate that the LSTM model yields the lowest prediction error, reflected by an RMSE value of 77.86, MAE of 58.23, and MAPE of 1.28%. Meanwhile, the hybrid CNN-LSTM model demonstrates more stable performance across different data proportions, achieving an RMSE of 75.71 and MAPE of 1.23% during the robustness test. The results indicate that LSTM is effective in capturing sequential dependencies inherent in financial time-series data, integrating convolutional feature extraction with sequential learning can improve prediction stability under varying training conditions. The results provide empirical insights into the selection of deep learning architectures for stock price prediction in the context of multi-sector holding companies.
The Impact of Parameter Scaling: Analysis of Specific Large Language Model Capabilities Putera, Ariya Uttama; Marcellino, Felix; Manalu, Sonya Rapinta; Muhamad , Keenan Ario
International Journal of Computer Science and Humanitarian AI Vol. 3 No. 1 (2026): IJCSHAI (In Press)
Publisher : Bina Nusantara University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21512/ijcshai.v3i1.15119

Abstract

Large Language Models (LLMs) are currently very diverse. Some of the largest include Chat-GPT, Gemini, Microsoft Copilot, Claude Sonet, Grok, and DeepSeek. Based on this, the plan of this research is to determine how efficient these AI models can be, based on their strengths in LLM training. In this study, we will examine the impact of LLM scaling parameters on the results of each local model we will test. This study also limits the number of parameters and classifies the questions to be asked. From these questions, we can identify and classify which local LLM models perform better when asked the same questions. Then, we will objectively evaluate each of them based on the results of the study. Thus, this study aims to establish a known correlation between scaling parameters and results. We also hope that it will be useful for improving work efficiency in selecting AI that suits user needs and expanding users' knowledge of AI so they can perform their jobs more efficiently and accurately. From this research, we conclude, aware of the results of the work that has been done, that local LLMs with large scaling are not entirely good and efficient. As with Gemma3, even with 12B parameters, the results weren't better than the Gemma3 model with 4B parameters. Alternatively, if you're using similar hardware to ours, you can use GPT-oss (openai/gpt-oss-20B) and Qwen3 (Qwen/Qwen3-4B & Qwen/Qwen3-8B), which offer good results in terms of reasoning and inference speed.
PoseTracker: Accuracy Evaluation of AI-Based Mobile Application for Exercise Posture Feedback Collhins, Billy; Mitta, Kalyana; Gunawan, Christian; Manalu, Sonya Rapinta
International Journal of Computer Science and Humanitarian AI Vol. 3 No. 1 (2026): IJCSHAI (In Press)
Publisher : Bina Nusantara University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21512/ijcshai.v3i1.15123

Abstract

In recent years, the rising of public health awareness has increased fitness activities participation. However, improper exercise form remains a significant contributor to injuries, particularly in unsupervised environments. To address this, PoseTracker’s accuracy was evaluated as a native Android application that provides real time feedback on exercise posture through MediaPipe based Human Pose Estimation (HPE) model. The system extracts 33 3D body landmarks, normalizes them to account for body scale, and employs cosine similarity to compare user movements against a reference dataset. Evaluations involving participants aged between 17 to 50 years old and 240 repetitions across four exercises demonstrated high detection accuracy: 88.33% for jumping jacks, 85% for squats, 83.33% for push-ups and 82% for sit ups. While performance can be influenced by environmental factors such as inconsistent lighting, camera positioning and incomplete body visibility, these results highlight the potential for lightweight, AI driven tools to support safe and self-guided fitness routines. Overall, the evaluations indicate that PoseTracker achieves reliable detection accuracy in distinguishing correct and incorrect exercise posture across multiple movement types under realistic conditions. Although performance variability exists due to environmental and system constraints, the accuracy levels observed demonstrate the feasibility of MediaPipe based Human Pose Estimation (HPE) for practical posture assessment in mobile fitness applications.
An End-to-End Architecture for Stock Market Prediction Integrating Mobile Application, Backend Services, and ML/DL Models Wilham, Abraham Kefas; William, William; Manalu, Sonya Rapinta
International Journal of Computer Science and Humanitarian AI Vol. 3 No. 1 (2026): IJCSHAI (In Press)
Publisher : Bina Nusantara University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21512/ijcshai.v3i1.15154

Abstract

Prior research on stock market prediction has predominantly focused on algorithmic accuracy, leaving a significant research gap in the system-level realization required for real-world delivery. This paper addresses this disparity by presenting an end-to-end stock prediction delivery system that operationalizes trained machine learning models within a mobile-centric architecture. Unlike model-centric studies limited to offline evaluation, this work focuses on the rarity of system-level implementation. Market data are periodically ingested into a managed relational database, where predictions are generated using a fixed historical window and persisted for downstream access. A cross-platform mobile application serves as the primary user interface, providing structured access to historical prices, predictions, and accuracy metrics via backend APIs without local model inference. A key novelty is the implementation of an in-memory caching layer to optimize responsiveness for repeated mobile access. Experimental results demonstrate that this architecture significantly improves efficiency, reducing average API response times by approximately 94% from 817 ms to 48,7778 ms compared to direct database queries. These findings underscore the critical role of mobile-oriented system design in bridging the gap between predictive modeling and practical deployment.
Analyzing Public Sentiment Toward the Makan Bergizi Gratis (MBG) Program on TikTok Using SVM and IndoBERT Winston, Alfredo; Darren, Nicholas; Lucky, Henry; Pradana, Rilo; Sagala, Noviyanti
International Journal of Computer Science and Humanitarian AI Vol. 3 No. 1 (2026): IJCSHAI (In Press)
Publisher : Bina Nusantara University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21512/ijcshai.v3i1.15184

Abstract

Social media has become a major platform for the public to express opinions toward government programs. This study analyzes public sentiment toward Indonesia’s Makan Bergizi Gratis (MBG) program using a text mining approach. A total of 11,730 TikTok comments related to the MBG program were collected and classified into positive, negative, and neutral sentiments. Two classification models were compared: a traditional Support Vector Machine (SVM) using TF-IDF features and a transformer-based model, IndoBERT. Experimental results show that IndoBERT outperforms the tuned SVM model, achieving an accuracy of 0.78 and a weighted F1-score of 0.78, compared to 0.73 accuracy and 0.73 F1-score obtained by the SVM. IndoBERT demonstrates better performance in handling neutral and context-dependent sentiments, indicating its effectiveness for analyzing Indonesian social media data related to public policy evaluation. This study contributes to the growing body of research on Indonesian sentiment analysis by providing an empirical comparison between classical machine learning and transformer-based models for analyzing public responses to government policies using social media data
Editorial, Foreword, and Table of Content Budiharto, Widodo
International Journal of Computer Science and Humanitarian AI Vol. 3 No. 1 (2026): IJCSHAI (In Press)
Publisher : Bina Nusantara University

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Page 3 of 3 | Total Record : 27

 1   2   3 

Filter by Year

2024 2026


Reset
Filter By Issues
All Issue Vol. 3 No. 1 (2026): IJCSHAI (In Press) Vol. 2 No. 2 (2025): IJCSHAI Vol. 2 No. 1 (2025): IJCSHAI Vol. 1 No. 1 (2024): IJCSHAI