Sentiment Analysis For User Review Classification On Jenius Application Using Naive Bayes Algorithm Based On Particle Swarm Optimization

Abstract

The rapid development of information technology and communication has facilitated various aspects of life, including the banking sector. Jenius is one of the digital banking applications that has been downloaded by five million users and continues to grow. With the increasing number of users, there is a growing number of opinions written about their experiences using the application, making it difficult to identify reviews through manual monitoring without textual data processing. This study aims to classify user reviews of the Jenius application on Google Playstore using the Naive Bayes algorithm and Particle Swarm Optimization feature selection. The data used consists of 3047 user reviews of the Jenius application collected from January 16, 2022 to April 13, 2023 and will be divided into two classes, namely positive and negative sentiment. This study also compares the Naive Bayes algorithm using PSO feature selection and without using PSO feature selection. The test results of the two methods indicate that the PSO feature selection with 800 iterations proves to be effective in optimizing the performance of the Naive Bayes algorithm model with an accuracy of 98.50%, precision of 97.81%, recall of 99.36%, and F1-score of 98.58%. Meanwhile, the performance level of the Naive Bayes algorithm without using PSO feature selection is lower with an accuracy of 96.68%, precision of 94.83%, recall of 99.04%, and F1-score of 96.88%.