Public Sentiment and Emotional Responses on X During Social Unrest in Indonesia: An Event-Based Analysis

Social media platforms play a critical role in shaping and reflecting public sentiment during socio-political crises. During periods of social unrest, X (formerly Twitter) serves as a primary channel for real-time public discourse. This study examines public sentiment and emotional dynamics expressed on social media during a social unrest event in Indonesia using an event-based computational approach. Tweets were collected across three temporal phases: pre-event, event-day, and post-event. Sentiment classification was performed using ensemble-based machine learning models, namely Random Forest and XGBoost, while emotion analysis employed a lexicon-based approach to identify dominant emotional expressions. Model performance was evaluated using standard classification metrics. The results indicate a strong dominance of negative sentiment, accounting for more than 95% of the analyzed tweets, with positive sentiment appearing marginally. Both classification models exhibit prediction patterns consistent with the highly imbalanced sentiment distribution. Emotion analysis reveals anger as the most dominant expressed emotion, while fear and sadness occur less frequently. These findings highlight the role of social media as a digital information infrastructure that amplifies collective emotional responses during social unrest, contributing to Information Systems research on crisis-related online public discourse.