<![CDATA[The growth of e-commerce in Indonesia has driven an increase in the volume of user reviews that openly express satisfaction and complaints. Manual analysis of large-scale reviews is inefficient, necessitating an automated computational approach. This study aims to classify user sentiment of the Blibli application based on 1,060 Google Play Store reviews using text mining and the Multinomial Naive Bayes algorithm with TF-IDF weighting, as well as to identify sentiment patterns per service aspect. Of the 350 manually labeled reviews, 245 were used for training and 105 for testing. Model evaluation results show an accuracy of 86.67%, precision of 82.54%, recall of 94.55%, and F1-score of 88.14%. The overall sentiment distribution across 1,060 reviews indicates 61.1% positive and 38.9% negative. Aspect-based analysis reveals that UI/UX is the only aspect with dominant positive sentiment (70.6%), while Application Performance (57.1%), Payment Features (63.9%), Customer Service (66.7%), and Delivery Experience (67.3%) are dominated by negative sentiment. These findings can serve as data-driven evaluation material for improving Blibli application services.]]>
Copyrights © 2026
