<![CDATA[The use of online shopping applications generates many reviews that record user experiences after using the service. In the Astro application, reviews on the Google Play Store can be processed into information about user rating trends without having to manually examine each comment. This research develops sentiment analysis on Astro reviews using Naïve Bayes. A total of 1,000 reviews were collected as a dataset, and then the text was normalized through case folding, tokenizing, stopword removal, and stemming. The preprocessed text is represented using TF-IDF before being used in the model training and testing process. The dataset is divided into 80% training data and 20% testing data. The testing resulted in an accuracy of 88.89%, with stronger performance on positive and negative sentiments compared to neutral ones. These findings indicate that TF-IDF processing and Naïve Bayes can be used to map the opinion trends of Astro users and assist in presenting evaluative information regarding the application services.]]>
Copyrights © 2026
