Sentiment Analysis on Fuel Purchase Policy Through MyPertamina Application Using NB and SVM Methods Optimized by PSO as Weight Optimation

Abstract

Sentiment analysis on the MyPertamina application can serve as a means to extract customer opinions about the application. This method involves collecting reviews from users who have utilized the MyPertamina application and classifying these reviews as positive or negative using sentiment analysis algorithms. After the reviews are classified, themes discussed in positive and negative reviews can be extracted, such as ease of use, payment speed, or technical issues. This provides a general overview of user expectations for the MyPertamina application and areas that may need improvement. Sentiment analysis of MyPertamina application comments using Naïve Bayes (NB) and Support Vector Machine (SVM) methods is a process to evaluate whether user comments on the MyPertamina application are positive or negative. NB and SVM are machine learning methods used to predict the category of an input based on given training data. In this study, user comments on the MyPertamina application are used as input and classified as positive, negative, or neutral based on previous training data. The goal of this sentiment analysis is to understand user perceptions of the MyPertamina application and enhance its quality. The research concludes that the implementation of data mining can assist in categorizing sentiments of MyPertamina reviews. The NB algorithm with the addition of Particle Swarm Optimization (PSO) proves to be the most effective method in this study compared to NB alone, SVM, and SVM + PSO. The NB algorithm with PSO optimization yields an accuracy of 79.49%, the highest precision of 79.57%, recall of 79.38%, and the highest AUC of 95.30%, falling into the category of excellent classification.