SENTIMENT ANALYSIS APLICATION RUANG GURU USING NAIVE BAYES AND SUPPORT VECTOR MACHINE

The rapid development of information technology has encouraged various innovations in the field of education, one of which is online learning applications such as Ruang Guru. As one of the largest learning platforms in Indonesia, Ruang Guru receives numerous user reviews that can be utilized to evaluate service quality. This study aims to perform sentiment analysis on user comments of the Ruang Guru application by comparing the performance of two popular classification algorithms: Naïve Bayes and Support Vector Machine (SVM). User comments were collected through web scraping from the Google Play Store and then went through a text pre-processing stage which included cleaning, case folding, tokenizing, stopword removal, and stemming. The data was then transformed into numerical representations using TF-IDF and Bag of Words feature extraction methods. The classification models were built using Multinomial Naïve Bayes and SVM with a linear kernel. The evaluation results show that the combination of SVM with TF-IDF produces higher accuracy compared to Naïve Bayes, achieving an accuracy level of more than 90% in crossvalidation. These findings reinforce the evidence that SVM performs better for high-dimensional text classification tasks such as sentiment analysis in the Indonesian language. This study is expected to provide valuable insights for Ruang Guru developers to improve service quality based on user opinions and serve as a reference for further research in the field of Natural Language Processing.