IMPLEMENTATION OF INDOBERT FOR PUBLIC SENTIMENT ANALYSIS TOWARD THE INDONESIAN GOVERNMENT’S FREE NUTRITIOUS MEAL PROGRAM

Abstract

The Free Nutritious Meal Program (MBG) is a government initiative aimed at improving public nutritional quality and reducing stunting rates in Indonesia. The implementation of this program has generated various public responses, particularly those expressed through social media platform X. Therefore, sentiment analysis is required to identify public opinion tendencies toward the program. This study aims to analyze public sentiment toward the Free Nutritious Meal Program using the Bidirectional Encoder Representations from Transformers (BERT) method, specifically the IndoBERT model. The dataset consists of 10,925 tweets collected through scraping using the X API and a public dataset from Kaggle. The data were processed through preprocessing stages including cleaning, case folding, and normalization. Sentiment labeling was conducted using a lexicon-based approach combined with manual labeling into three sentiment classes, positive, negative, and neutral. To address data imbalance, random oversampling was applied prior to model training. Model performance was evaluated using a confusion matrix with accuracy, precision, recall, and F1-score metrics. The results show that the IndoBERT model performs well in classifying public sentiment, achieving an accuracy of 90.2%, precision of 90.4%, recall of 90.2%, and an F1-score of 90.3%. Overall, public sentiment toward the Free Nutritious Meal Program is predominantly positive.