<![CDATA[– The rapid growth of e-commerce in Indonesia has made user reviews a critical source of feedback, yet discrepancies between star ratings and actual sentiment often mislead businesses. This study employs the Multinomial Naïve Bayes algorithm to analyze sentiment in 25,000 Shopee application reviews collected via web scraping. The research utilizes TF-IDF for feature extraction and Bigram analysis to capture contextual meaning, addressing the challenge of imbalanced data (82% positive, 18% negative). The objective is to accurately classify user sentiment into positive and negative categories to provide actionable insights beyond numerical ratings. The model achieved a classification accuracy of 91.96%, with a high Recall of 77% for the minority negative class, ensuring effective identification of user complaints. Bigram analysis revealed that "delivery speed" is the primary driver for both satisfaction and dissatisfaction. The study confirms that Naïve Bayes is a robust and scalable solution for large-scale sentiment analysis in the Indonesian e-commerce context, offering a reliable tool for business intelligence]]>
Copyrights © 2026
