Klasifikasi Opini Publik terhadap Kenaikan PPN 12% di Platform X menggunakan Multinomial Naïve Bayes

Abstract

The increase in Value-Added Tax to 12% in 2025 has sparked diverse public opinions on the social media platform X (Twitter). This study aims to classify public sentiment toward the policy using Multinomial Naïve Bayes with a Term Frequency-Inverse Document Frequency (TF-IDF) approach. Multinomial Naïve Bayes is a probabilistic classification algorithm that assumes feature independence. Data were collected through web crawling using the keyword "ppn 12%" and underwent pre-processing, including text normalization, stopword removal, and stemming. To address class imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) was applied. The best-performing model was obtained by tuning the alpha hyperparameter to 0.01, achieving an average accuracy of 83.37%, precision of 83.32%, recall of 83.38%, and an F1-score of 82.99% using 10-fold cross-validation. The findings indicate that Multinomial Naïve Bayes, combined with SMOTE and hyperparameter tuning, effectively classifies public sentiment and provides insights into public responses regarding the Value-Added Tax policy.