Sentiment Analysis of the Free Nutritious Meal Program Using IndoBERT and RCNN Methods

Abstract

This study examines public sentiment regarding the Free Nutritious Meal Program through a deep learning-based sentiment classification methodology applied to X and TikTok. The suggested method uses a hybrid IndoBERT RCNN architecture, with IndoBERT being used to extract features and RCNN being used to classify sentiment. There are 10,000 comments from each platform in the dataset. These comments went through preprocessing and sentiment labeling steps. Model evaluation was conducted using stratified K-fold cross-validation with different combinations of learning rate, batch size, and epochs. The best configuration achieved an accuracy and F1-score of 78% on X and 83% on TikTok. The model performs well in identifying overall sentiment patterns, although neutral sentiment remains challenging to classify, particularly in X data containing sarcastic or indirect language. These findings provide empirical insights into cross-platform sentiment characteristics and highlight the potential of this approach for testing sentiment monitoring strategies across platforms.