OPTIMIZATION OF SENTIMENT ANALYSIS ALGORITHM ON YOUTUBE MUSIC APPLICATION WITH COMPARISON OF NAIVE BAYES AND SUPPORT VECTOR MACHINE

Abstract

This study conducted a sentiment analysis of a new application created by YouTube, namely the YouTube Music application. Sentiment analysis is used to understand user opinions about a service or product. In conducting sentiment analysis, researchers compared the Naïve Bayes and Support Vector Machine algorithm methods, which were then optimized using Particle Swarm Optimization for both algorithms. Research data was collected through web scraping from the Google Play Store which contained reviews from YouTube Music application users. Each review was labeled with positive and negative sentiment based on the context and emotions contained therein. The results showed that the Naïve Bayes algorithm model had a higher accuracy rate of 87.17% and the Support Vector Machine had an accuracy of 91.80%. The Particle Swarm Optimization method successfully optimized the evaluation process using the Confusion Matrix, with the initial Naïve Bayes accuracy of 87.17% to 91.80%, the initial accuracy of the Support Vector Machine of 85.20% to 85.51%. The results of sentiment analysis using Naïve Bayes with Particle Swarm Optimization on the YouTube Music application show that users responded positively with a total of 5,967 positive sentiments and 1,657 negative sentiments.