Twitter Sentiment Analysis of Electric Vehicle Subsidy Policy using Naïve Bayes Algorithm

Abstract

This research aims to apply the Naïve Bayes classifier in Indonesian-language sentiment analysis, regarding electric vehicle subsidy policies using Twitter data with the query 'subsidi kendaraan listrik'. The stages of analysis include data pre-processing, tokenization, stemming, forming the Naïve Bayes model, and evaluating model performance using accuracy, precision, and recall. The SMOTE technique is used to deal with class imbalances, in which the majority of negative sentiments towards the policy is 66%. The results obtained from the 10-fold Cross Validation with the binary classification (positive and negative sentiment) show that the accuracy value of the model is 69.49%, with precision and recall values of 53.27% and 74.26%.