Penerapan Multinomial Naïve Bayes dan Chi-Square pada Analisis Sentimen Makan Bergizi Gratis

Abstract

The Free Nutritious Meal Program (MBG) is a policy initiated by the government in order to create a healthy superior generation. This policy has generated various responses from the public, both positive, negative, and neutral. This study aims to analyze public sentiment towards the program by utilizing 2.165 comments from YouTube. Data were analyzed using the Multinomial Naïve Bayes algorithm after going through the labeling, text preprocessing, and TF-IDF representation stages. Feature selection was carried out with 10 threshold values ​​of the Chi-Square method to evaluate the effect of the number of features on model performance. Hyperparameter tuning with alpha and fit_prior was carried out through Grid Search with 5-fold cross validation. The results showed that the best performance was achieved when the threshold Chi-Square was 100% or when all features were used, which resulted in an accuracy of 83.48% and a macro F1 score of 83.42%. Visualization of the sentiment distribution shows the dominance of negative sentiment at 35.6%, followed by positive at 34.9%, and neutral at 29.5%, indicating public dissatisfaction with the policy.