<![CDATA[The Gojek application has become one of the largest platforms providing transportation, food delivery, and digital payment services in Indonesia. Every day, thousands of users provide reviews in the form of criticisms, complaints, and compliments through the Google Play Store. However, the massive volume and unstructured nature of these reviews pose challenges for management in monitoring user satisfaction manually, objectively, and rapidly. The primary issue in analyzing this review data is the high level of class imbalance between the number of positive and negative reviews, where positive reviews frequently dominate significantly. This data imbalance becomes crucial as it tends to bias standard classification models and reduce their predictive accuracy toward the minority class (negative reviews), even though these negative reviews contain vital complaints necessary for system improvement. Therefore, this study aims to analyze the sentiment of Gojek users by comparing the performance of the Multinomial Naïve Bayes (MNB) and Logistic Regression (LR) classification algorithms, while simultaneously addressing the data imbalance issue. The research process encompasses text preprocessing stages, including cleansing, case folding, stopword removal, and stemming. Feature extraction is performed using Term Frequency-Inverse Document Frequency (TF-IDF) and Count Vectorizer methods based on Unigram and Bigram schemes. Data splitting utilizes a proportion of 80% training data and 20% testing data, where the Synthetic Minority Over-sampling Technique (SMOTE) is applied specifically to the training data to resolve the text category imbalance. The evaluation results are measured based on accuracy, precision, recall, and F1-score metrics. This research is expected to provide recommendations for the best algorithm for large-scale review text classification, as well as data-driven insights for Gojek developers to enhance service quality based on genuine user sentiments.]]>
