Sentiment Analysis of Bapenda South Sulawesi Mobile Application on Google Play Store Using Support Vector Machine

Abstract

This study analyzes user sentiment toward the Bapenda Sulsel Mobile application, an e-government platform developed by the Regional Revenue Agency of South Sulawesi, Indonesia. The research aims to evaluate user feedback and identify areas for improvement to enhance user satisfaction. Using sentiment analysis, user reviews from Google Play Store were collected and classified into positive, negative, and neutral sentiments through the Support Vector Machine (SVM) algorithm. Preprocessing steps such as tokenization, stopword removal, and stemming were applied to prepare the data. Term Frequency-Inverse Document Frequency (TF-IDF) was used for feature extraction to enhance classification accuracy. The SVM model demonstrated an overall accuracy of 80%, achieving a high recall of 98% for positive reviews but only 40% for negative reviews, reflecting challenges in handling class imbalance. Results show that 72% of users expressed positive sentiment, praising the app’s functionality and ease of use. However, 28% of reviews were negative, citing issues like technical bugs and usability challenges The findings highlight the app’s strengths in delivering e-government services and its role in improving tax management. However, the significant proportion of negative feedback emphasizes the need for addressing user concerns. Recommendations include balancing the dataset, refining the SVM model, and prioritizing improvements based on user feedback. This study contributes to the broader understanding of applying sentiment analysis in evaluating e-government platforms and offers actionable insights for enhancing the user experience.