<![CDATA[The rapid growth of Instagram user reviews on the Google Play Store poses challenges in understanding sentiment quickly and accurately. This research aims to develop an automatic sentiment analysis dashboard based on the Indonesian-RoBERTa model. Review data was collected using the google_play_scraper library and analyzed using a fine-tuned model. Fine-tuning was performed on the w11wo/indonesian-roberta-base-sentiment-classifier model using Indonesian tweet datasets during the PPKM period with 23,645 labeled data points (positive, neutral, negative). The preprocessing process included text cleaning, tokenization, and class weighting. Model evaluation used precision, recall, F1-score, and confusion matrix metrics. Test results showed good performance on positive and neutral classes, but performance on the negative class still needs improvement. The dashboard successfully performed scraping, sentiment prediction, and data visualization automatically. This research demonstrates the potential application of transformer-based models in Indonesian language and supports data-driven decision making.]]>
Copyrights © 2025
