Sentiment Analysis of TikTok Comments on Indonesian Presidential Elections Using IndoBERT

TikTok, a popular social media platform, hosts numerous trending topics, including discussions about the Indonesian presidential and vice-presidential elections (Pilpres). Netizens frequently share their views during this period, resulting in a mix of positive, neutral, and negative comments that can significantly influence public opinion. This research aims to develop a model for classifying Indonesian comments on TikTok using the pre-trained IndoBERT model. The dataset comprises 36,991 comments: 10,600 positive, 10,136 neutral, and 16,255 negative. The research process involves data preprocessing, labeling, training, validation, and testing. The developed model, named Indonesia-Pemilu-Sentiment-Classification, is hosted on Hugging Face. Testing results show an accuracy of 92.1%, a precision of 92.3%, a recall of 92.1%, and an F1-score of 92.1%, indicating the model's high accuracy in classifying sentiment in comments. The model demonstrates high performance and reliability in sentiment classification, contributing to a better understanding of public opinion during the Indonesian presidential elections.