Comparing Naïve Bayes and SVM for Public Sentiment on Online Gambling Policies

Abstract

Online gambling has become a major digital and social issue in Indonesia, prompting the government to implement various enforcement policies such as website blocking, account suspension, and advertisement filtering. However, illegal gambling content continues to circulate widely on social media platforms, generating diverse public reactions. This study aims to analyze public sentiment toward online gambling policies in Indonesia using Natural Language Processing (NLP) techniques and compare the performance of Naïve Bayes and Support Vector Machine (SVM) algorithms. Data were collected from Twitter (X), Facebook, and Instagram, resulting in 1,110 social media posts. The dataset was processed through text preprocessing, stemming, stopword removal, and TF-IDF feature extraction. To address class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) was applied before model training. The experimental results show that SVM achieved better performance with an accuracy of 91.84%, while Naive Bayes achieved 86.39%. WordCloud and confusion matrix analyses indicate that negative sentiment is dominated by dissatisfaction with slow enforcement and persistent gambling advertisements, whereas positive sentiment reflects public support for stricter regulation. These findings demonstrate that SVM is more effective for Indonesian-language sentiment classification and can support data-driven evaluation of digital governance policies.