Sentiment Analysis on KPU Performance Post-2024 Election via YouTube Comments Using BERT

Abstract

This research aims to analyze public sentiment regarding the performance of the General Election Commission after the 2024 presidential election using the BERT (Bidirectional Encoder Representations from Transformers) model. Given the General Election Commission's crucial role in maintaining election integrity and the importance of transparency in Indonesian democracy, understanding public opinion through sentiment analysis is essential. Data was collected from YouTube comments, a platform increasingly popular for public expression. The analysis process began with data preprocessing, including case folding, text cleaning, tokenization, and stop word removal. The BERT model was then applied to classify the sentiment of the comments, with the model's performance evaluated using 10-fold cross-validation. The evaluation results showed that the first fold (k=1) achieved the best performance with an accuracy of 96%, precision of 96%, recall of 96%, and an F1-score of 96%, indicating the model's effectiveness in accurately classifying sentiment. In contrast, the ninth fold (k=9) exhibited the lowest accuracy at 86% with other metrics also lower, suggesting performance instability potentially caused by data variability. Accuracy and loss graphs confirmed that the first fold experienced consistent accuracy improvements and significant loss reduction, while the ninth fold showed performance fluctuations. This study provides valuable insights into public sentiment regarding the General Election Commission performance, with BERT demonstrating significant potential for sentiment analysis on social media platforms like YouTube.

GS Cited Analysis