<![CDATA[The growth of e-commerce has transformed consumer behavior, with Shopee emerging as one of the leading platforms in Southeast Asia and particularly dominant in Indonesia. Millions of user reviews on the Google Play Store capture diverse experiences, yet their unstructured nature hinders efficient extraction of actionable insights. This study addresses the challenge by developing an automated sentiment analysis system for Shopee user reviews, focusing on the effective use of the Naïve Bayes algorithm for Indonesian-language data. While Naïve Bayes is widely applied in text classification, this research distinguishes itself by integrating rigorous preprocessing tailored to colloquial and context-specific Indonesian app reviews, coupled with TF-IDF weighting, to enhance classification performance. A dataset of 4,000 reviews was collected via web scraping, labeled automatically based on user ratings, and split into 80% training and 20% testing subsets. Preprocessing included cleaning, case folding, tokenization, and stemming to standardize textual input. The proposed model achieved an accuracy of 83%, precision of 81%, recall of 90%, and F1-score of 85%, indicating strong performance despite class imbalance and the prevalence of ambiguous or sarcastic expressions. The results demonstrate that a lightweight probabilistic classifier, when combined with domain-specific preprocessing, can yield competitive accuracy while maintaining computational efficiency. This study contributes to sentiment analysis research in underrepresented linguistic contexts and offers a practical framework for e-commerce platforms to systematically interpret large-scale user feedback, prioritize feature improvements, and enhance customer satisfaction strategies.]]>
