Analisis Sentimen Kebijakan Insentif Mobil Listrik Menggunakan TF-IDF dan Naïve Bayes

Abstract

This study aims to analyze public sentiment regarding the Indonesian government’s electric vehicle (EV) incentive policy using YouTube comments as the data source. The research applies text preprocessing steps including cleaning, normalization, stopword removal, tokenization, and stemming to prepare the textual data. The cleaned data is transformed into numerical representation using the Term Frequency-Inverse Document Frequency (TF-IDF) method and classified using the Multinomial Naïve Bayes algorithm. To address class imbalance in the dataset, Synthetic Minority Over-sampling Techique (SMOTE) is applied. The model evaluation metrics include accuracy, precision, recall, and F1-score. Based on the evaluation, the model achieves an accuracy of 71%. The model performs better in classifying negative comments, as shown by a higher recall and F1-score in the negative class compared to the positive class. These findings indicate that public responses to the EV incentive policy tend to be more critical. This study provides insights into public opinion that can serve as a valuable reference for policymakers in designing more effective and well-communicated incentive strategies for promoting electric vehicle adoption in Indonesia.