Comparison of Naïve Bayes and Random Forest on Sentiment Analysis of Brand Reputation Provider Live. On

Internet users in Indonesia continue to increase yearly, reaching 215.63 million people in 2022. However, many internet users in Indonesia complain about the services provided by internet service providers. This complaint arises because the service provider does not provide full control over the use of data services and the distribution of quotas that are not appropriate. To solve this problem, Live. On is here as a new digital service provider to improve the user experience. To evaluate the success of Live. On, a sentiment analysis was carried out through the Google Play platform to see user feedback regarding the services provided. In this study, data was collected through a data scraping process on Google Play. Furthermore, the data were analyzed using the Naïve Bayes and Random Forest classification methods. The Net Reputation Score (NRS) calculation is used to measure the level of user satisfaction. The comparison results show that when the SMOTE technique is not used, Random Forest performs better than Naïve Bayes. Random Forest has 80.37% accuracy, 61% precision, 58% recall, and 57% f1-score, while Naïve Bayes has 80% accuracy, 54% precision, 58% recall, and 56% f1-score. When the SMOTE technique is applied, Random Forest still performs better with 78% accuracy, 79% precision, 78% recall, and 78% f1-score, while Naïve Bayes has the same performance on all metrics with 73%. In addition, based on NRS calculations using Naïve Bayes prediction results, Live. On provider reputation has a value of −33.08%, while using Random Forest prediction results, its reputation reaches −41.08%. Both reputation scores are below average, so Live. On is considered to have a poor reputation. Therefore, Live. On needs to fix complaints felt by users to improve its reputation.