<![CDATA[The advancement of the digital era has driven increased usage of online reservation applications, including Traveloka. The abundance of user feedback available on the Google Play Store platform has the potential to become a valuable database for development teams in improving service quality. However, the characteristics of unstructured and spontaneous reviews pose challenges in conventional data processing.This research aims to explore sentiment in Traveloka application user comments using the Multinomial Naïve Bayes algorithm. The dataset used consists of 1,500 review samples obtained through web scraping techniques from the Google Play Store. The research methodology includes several data preprocessing stages, including data cleaning, case normalization, word tokenization (tokenizing), stopword removal, and word stemming to their base forms (stemming). Subsequent processes include data categorization, feature extraction using the Term Frequency–Inverse Document Frequency (TF-IDF) approach, and building a classification model with the Multinomial Naïve Bayes algorithm.Test results show that the model is capable of classifying sentiment with an accuracy rate of 79%. The model demonstrates high recall values in identifying negative reviews (0.97), but the recall value for positive reviews remains limited (0.64). This indicates that the model has higher sensitivity to negative expressions. From a total of 1,500 review data, there were 461 positive reviews and 543 negative reviews that were successfully categorized clearly.The findings in this study prove that the implementation of the Multinomial Naïve Bayes algorithm is quite efficient in sentiment classification of user reviews, and is capable of providing strategic insights that can be utilized by development teams to improve application service quality]]>
