<![CDATA[This research focuses on identifying user sentiment toward the TJ: Transjakarta application based on reviews available on the Google Play Store. A total of 3,431 user reviews were collected using web scraping techniques. The collected data were processed through several preprocessing stages, including text cleaning, normalization, tokenization, stopword removal, and stemming. Sentiment labels were initially assigned using a dictionary-based lexicon approach with the InSet lexicon, resulting in 1,991 positive and 1,440 negative reviews. The dataset was subsequently divided into training, validation, and testing subsets before fine-tuning the IndoBERT-base-p1 model. Experimental results show that the proposed model achieved an accuracy of 87.76%, with balanced precision and recall across both sentiment classes. Further analysis using word cloud visualization indicates that positive sentiment is mainly related to route accessibility and schedule information, while negative sentiment is dominated by technical issues and transaction-related problems. These results confirm the effectiveness of IndoBERT for Indonesian sentiment analysis and its potential use in evaluating the quality of public transportation applications]]>
Copyrights © 2026
