SENTIMENT ANALYSIS OF PLN MOBILE APPLICATION SERVICES USING NAIVE BAYES, SUPPORT VECTOR MACHINE (SVM) AND DECISION TREE METHODS

Abstract

The advancement of information technology has driven public service providers such as PLN to introduce digital innovations, one of which is the PLN Mobile application that enables customers to access various services online. As the number of users increases, numerous reviews have been submitted through the Google Play Store platform, which can be utilized to evaluate service quality. This study aims to conduct sentiment analysis on user reviews of the PLN Mobile application using three classification algorithms: Naïve Bayes, Support Vector Machine (SVM), and Decision Tree. A total of 4,992 review data were collected and processed through text preprocessing stages, including case folding, tokenization, stopword removal, stemming, and vectorization using the TF-IDF method. The data were then split into training and testing sets with a ratio of 80:20 and trained using the three classification algorithms. Model evaluation was conducted using precision, recall, f1-score, and accuracy metrics. The evaluation results indicate that the SVM algorithm delivers the best performance with an accuracy of 94%, followed by Naïve Bayes and Decision Tree, each with an accuracy of 91%. However, all three models demonstrated limited effectiveness in detecting neutral sentiments. Based on these findings, the SVM algorithm is recommended as the most effective model for sentiment classification of PLN Mobile application reviews.