Sentiment Analysis of Students Toward Campus Services Using Indobert and Explainable AI Based on Logistic Regression

Abstract

This study aims to analyze student sentiment toward campus services using IndoBERT, Logistic Regression, and Explainable Artificial Intelligence (XAI) based on SHAP. The dataset consists of 100 student comments covering academic services, administrative services, information systems, campus facilities, library services, and student services. Each comment is classified into positive, neutral, and negative sentiment categories. The research stages include data collection, sentiment labeling, text preprocessing, contextual representation using IndoBERT, classification using Logistic Regression, model evaluation, and interpretation using SHAP. The results show that positive sentiment represents 45% of the dataset, followed by negative sentiment at 35% and neutral sentiment at 20%. The classification model achieves an accuracy of 85.00%, with precision, recall, and F1-score of 87.00%. The information system aspect has the highest number of negative comments, indicating the need for improvement in system performance and accessibility. SHAP analysis identifies words related to speed, friendliness, and helpfulness as important contributors to positive sentiment, while slow response, errors, and difficulties contribute to negative sentiment. The proposed approach provides both accurate sentiment classification and interpretable information to support campus service improvement.