Analisis Sentimen pada Sengketa Pilpres 2024 dengan Multinomial Naïve Bayes dan Chi-Square

Abstract

The Constitutional Court officially rejected all appeals against the 2024 Presidential Election results in Jakarta, drawing public attention. People expressed their opinions through social media platforms, including YouTube. Therefore, this research was conducted to analyze public sentiment regarding this political phenomenon. This study uses primary data on comments from 3 YouTube videos that were crawled with a total of 4103 data. The data underwent labeling and then went through 5 stages of text pre-processing. Subsequently, modeling was performed using the Multinomial Naïve Bayes Algorithm with testing on 10 variations of the Chi-Square threshold, followed by validation with 10-fold cross-validation. The research yielded an optimal model performance, including accuracy, precision, recall, and F1-score, stabilizing at 72% with a Chi-Square feature selection of 70% and 90%. Without feature selection, the model's performance was at 71.5%. Model validation with the highest and average K-Fold Cross-Validation scores were sequentially at 74% and 69%. The built model is sufficiently good at analyzing sentiment and provides insights that 52% of Indonesians are dissatisfied with the Constitutional Court's decision based on YouTube comments.