The Free Nutritious Meals (MBG) program is a government program aimed at supporting nutritional fulfillment and improving human resource quality. The high level of public interaction on X makes the platform a potential source of data for identifying public responses to the implementation of MBG. This study aims to analyze public sentiment toward the MBG program and compare the performance of the Naïve Bayes and Support Vector Machine (SVM) algorithms. The research data consisted of 5,048 Indonesian-language tweets collected through Browser Automation Scraping. The data were processed through cleaning, case folding, normalization, tokenizing, stopword removal, and stemming, followed by sentiment labeling into positive, neutral, and negative categories using a lexicon-based approach. The dataset was divided using an 80:20 ratio into 3,810 training data and 953 test data. SMOTE was applied to the training data to address class imbalance. The results showed that Naïve Bayes achieved an accuracy of 35.36% before SMOTE and increased to 70.20% after SMOTE. SVM with SMOTE achieved an accuracy of 85.73%, correctly classifying 817 of 953 test data, which was 15.53 percentage points higher than Naïve Bayes with SMOTE. These results indicate a difference in the performance of the two algorithms, with SVM achieving higher overall classification performance on the research dataset.