SENTIMENT ANALYSIS MODEL ON ELECTRIC VEHICLES USING INDOBERTWEET AND INDOBERT ALGORITHM

The increasing adoption of electric vehicles in Indonesia has sparked various public opinions, necessitating sentiment analysis to understand societal perspectives. This study aims to compare the performance of two transformer-based models, IndoBERTweet and IndoBERT, in analyzing sentiments towards electric vehicles in Indonesia. Using a dataset collected from Indonesian language tweets and online comments, the data undergoes preprocessing, sentiment labelling into positive, negative, and neutral sentiments, and subsequent fine-tuning of both models. The models are evaluated based on accuracy, precision, recall, and F1-score. Experimental results demonstrate that IndoBERTweet achieves superior performance compared to IndoBERT in sentiment classification. The best performance recorded for IndoBERTweet was an accuracy of 82,40%, with an F1-score of 82,39%, while IndoBERT achieved an accuracy of 75,98% and an F1-score of 75,46%. These findings highlight the importance of using domain-spesific models for sentiment analysis and contribute to advancements in Indonesia-language natural language processing (NLP).