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

Abstract

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 cross
validation. 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.