Analisis Sentimen Masyarakat terhadap Program Makan Bergizi Gratis di Twitter Menggunakan Metode Support Vector Machine

Abstract

Twitter (X) social media has become a primary platform for the public to express opinions on public policies, including the Free Nutritious Meal Program (MBG) launched by the Indonesian government. This study aims to analyze public sentiment towards the MBG program using the Support Vector Machine (SVM) method based on TF-IDF weighting. Data were collected through crawling on Twitter using keywords "MBG", "makan bergizi gratis", "program makan bergizi gratis", and "embege" from April 2024 to April 2026, yielding 10,043 tweets. Preprocessing was carried out through cleaning, case folding, tokenizing, stopword removal, and stemming using the Sastrawi library. Buzzer detection was applied to filter 53 abnormal accounts (0.53%). Sentiment labeling using a lexicon-based approach resulted in 69.21% positive sentiment, 28.30% neutral, and 2.49% negative. TF-IDF weighting produced 5,000 features. The SVM model was optimized using GridSearchCV with parameters C: 0.1-100, gamma: 0.001-1, kernel: linear and RBF. The tuning results showed the best parameters C=10, gamma=1, kernel=linear with a cross-validation accuracy of 99.66%. Model evaluation on the test data yielded an accuracy of 99.90%, precision of 99.90%, recall of 99.90%, and F1-score of 99.90%. These results prove that the SVM method is effective for sentiment classification of the MBG program and can be recommended as a tool for evaluating public policy based on public opinion.