Implementation of Machine Learning in Sentiment Analysis of The MyTelkomsel Application Using Google Playstore Review Data

Abstract

The MyTelkomsel application is a digital access platform that provides telecommunications services. Therefore, sentiment analysis of MyTelkomsel application users is relevant for obtaining valuable insights for application development and management. This research aims to conduct sentiment analysis and compare methods on review data of the MyTelkomsel application. The dataset used is divided into two topics: service and user reviews. The labeling method in this research uses Lexicon Based and Indonesian Language Experts with three classes: positive, negative, and neutral. The labeled review dataset is then applied with SVM and Random Forest methods. The results obtained from applying two datasets with two labeling approaches indicate that the approach by experts tends to be more accurate compared to the lexicon-based approach because the highest accuracy of the lexicon-based approach is 79%, while the expert labeling achieves an accuracy of 83%. Additionally, in this study, the SVM algorithm demonstrates the highest accuracy, namely 83%, on the user dataset analyzed by Indonesian Language Experts.