Comparative Analysis of Naive Bayes and Support Vector Machine for Sentiment Analysis of Mobile Banking Application Reviews in Indonesia

Abstract

This study aims to analyze user sentiment toward mobile banking applications in Indonesia using machine learning approaches and to compare the performance of Naive Bayes and Support Vector Machine (SVM) algorithms. The dataset was collected through scraping user reviews from Google Play Store, focusing on three major mobile banking applications: BCA Mobile, Livin by Mandiri, and BRImo. The research methodology includes data preprocessing (case folding, cleaning, and stemming), feature extraction using Term Frequency–Inverse Document Frequency (TF-IDF), optional feature selection using Chi-Square, and classification using Naive Bayes and SVM. The performance of each model was evaluated using accuracy, precision, recall, and F1-score metrics. The experimental results show that the SVM algorithm achieved the highest accuracy of 89.80%, outperforming Naive Bayes, which achieved 86.22%. Meanwhile, the application of Chi-Square feature selection did not significantly improve model performance and slightly reduced accuracy to 85.71%. These findings indicate that SVM is more effective in handling high-dimensional text data for sentiment classification tasks, while TF-IDF representation alone is sufficient without additional feature selection. This study contributes to the evaluation of machine learning methods for sentiment analysis in mobile banking applications and provides insights for improving digital banking services based on user feedback.