Sentiment Analysis of BPD DIY Mobile Banking Application Using SVM and KNN Methods

Abstract

This study aims to conduct sentiment analysis on user reviews of the BPD DIY Mobile Banking application available on the Google Play Store. The analysis is crucial due to the increasing number of user complaints regarding technical performance and user experience that have not been systematically addressed. Two machine learning algorithms, the Support Vector Machine (SVM) and the K-Nearest Neighbour (KNN), were used to classify reviews into positive and negative sentiments.  The dataset comprises 1,211 user reviews collected through web scraping and processed with comprehensive preprocessing stages, including cleaning, tokenizing, case folding, stopword removal, normalization, and stemming. The novelty of this research lies in the integration of Indonesian-specific preprocessing techniques and a comparative evaluation of two classification models, which are rarely applied in similar studies focused on regional banking applications.  The results indicate that SVM outperforms KNN, achieving 81.48% accuracy, 82.30% precision, and 88.50% recall, while KNN only reaches 55.56% accuracy, 63.00% precision, and 65.50% recall. With this level of accuracy, the SVM-based model can be effectively utilized for real-time sentiment monitoring and to identify critical issues in user experience. These findings offer strategic insights for BPD DIY to enhance application quality, particularly in addressing technical problems frequently highlighted by users.