Public Opinion Sentiment Analysis Towards Government Budget Efficiency Policy on Twitter (X) Using the Naïve Bayes Classifier Algorithm

The government’s budget efficiency program, mandated through Presidential Instruction No. 1 of 2025, represents a strategic initiative to maximize the effectiveness of national (APBN) and regional (APBD) spending while minimizing waste. This policy has triggered diverse public responses, particularly on Twitter (X), which serves as one of the most widely used platforms in Indonesia for expressing opinions openly. This study investigates public sentiment toward the policy by applying the Multinomial Naïve Bayes Classifier algorithm. A total of 1,000 tweets were collected through crawling between January and March 2025 using the keywords “government budget efficiency” and “APBN savings.” The analytical process involved several steps, including text preprocessing, automatic labeling with the Indonesian InSet lexicon-based dictionary, and TF-IDF weighting. The dataset was divided into 80% training data and 20% testing data. Labeling results identified 703 positive tweets and 297 negative tweets. Model performance evaluation using a confusion matrix achieved an accuracy of 77%, precision of 57.14%, recall of 82.76%, and an F1-score of 67.6%. Although this study focuses only on binary sentiment classification (positive and negative), the findings demonstrate that the proposed method is sufficiently effective in classifying public sentiment related to the government’s budget efficiency policy. The results also provide significant insights into public opinion and can serve as a reference for policymakers as well as for future research on social media-based sentiment analysis.