Sentiment Analysis and Topic Detection of Mobile Banking Application Review

Abstract

Understanding user needs and application quality are difficult things in developing an application. Sentiment analysis and topic modeling based on application review can be used to understand the user needs and application quality. This research aimed to determine customer sentiment towards mobile banking applications and what aspects need to be improved or maintained from the application. The data come from application user reviews on Google Play Store, which amounted to 6194 data. Labeling is done manually, generates two main classes namely positive and negative classes. The sentiment analysis process is done using Naive Bayes models. While the topic modeling process is carried out using the LDA algorithm. The results of the experiment were Naive Bayes method has a good level of accuracy, recall, and precision. The highest accuracy, recall, and precision are at the value of k=5, which is 86.762% accuracy, 93.474% for recall, and 92.482% for precision. Based on the LDA algorithm, the most frequent topics in negative classes are related to OTP code delivery constraints, application login problems, and network connection. On the other hand, the most frequent topics in positives classes included ease, simplicity, and helpfulness.