Journal on Informatics Visualization and Social Computing

Development of an Eye Care Chatbot Based on Llama 3.2 1B Using 4-bit QLoRA Technique

2026-08-07T09:54:00+07:00

Background: The ratio of ophthalmologists to the population in Indonesia is extremely low, leading to unequal access to eye care and making it difficult for the public to obtain early information regarding eye symptoms.
Aims: This study aims to develop and validate an efficient, domain-specific eye care chatbot using the Llama 3.2 1B Instruct model, focusing on an optimal fine-tuning pipeline for limited hardware constraints.
Methods: Utilizing an Experimental Research (Model Development) design, a QLoRA 4-bit quantization technique was applied on a Google Colab T4 GPU (15.64GB VRAM). The training dataset comprised 16,742 samples from Kaggle Eye Care and MedQuad Indonesian Translation, with a subset of 5,000 samples used for experiments.
Result: After training for 3 epochs (750 steps), the Training Loss decreased from 1.3394 to 0.7188 (46.3% improvement), while Perplexity reached 2.3620, categorized as excellent. The model maintained stability with a final gap of 0.1407 between training and validation loss.
Conclusion: The Meta-Llama 3.2 1B Instruct model, fine-tuned with 4-bit QLoRA, is highly effective for building a domain-specific healthcare chatbot, successfully operating within an 8GB VRAM limit while maintaining high generalization capabilities.

A Web-Based Forward-Chaining Expert System for Smartphone Selection A Decision Support Tool for Novice

2026-08-10T09:35:00+07:00

Spatial Analysis and Classification of Traffic Congestion Levels Using a Hybrid Machine Learning Approach in Badung Regency, Bali

2026-08-10T09:36:47+07:00

Background: Traffic congestion in Badung Regency, Bali, presents complex challenges for transportation management in a strategic tourism area, necessitating intelligent monitoring methods that surpass conventional surveys.
Aim: This study develops a Hybrid Machine Learning framework that integrates unsupervised and supervised learning techniques to map spatial patterns and precisely predict congestion anomalies.
Methods: Initially, the K-Means Clustering algorithm validated the segmentation of traffic data into five optimal clusters (k=5) based on Elbow and Silhouette Score tests, which subsequently served as the ground truth for data labeling. Furthermore, a comparative evaluation of Random Forest, XGBoost, and Gradient Boosting models was conducted to determine the best predictive performance.
Result: Experimental results identified Random Forest with strict pruning parameters (max_depth=3) as the optimal model, balancing accuracy and generalization stability while effectively avoiding overfitting through learning curve analysis. This model achieved a testing accuracy of 92.00% and a cross-validation score of 89.69%, with the Travel Time Index (TTI) and average speed identified as the primary determinants.
Conclusion: Spatial visualization demonstrated high consistency between the model's prediction maps and actual field conditions, confirming that this approach is effective as a foundation for an Early Warning System to support data-driven traffic management policies in Bali.