Analisis Sentimen Program Makan Bergizi Gratis pada Podcast Bocor Alus Politik dengan Algoritma Naive Bayes

Abstract

The Free Nutritious Meal (MBG) program initiated by the elected administration has evolved into a strategic public policy, yet it has garnered diverse responses from various strata of society. These opinion dynamics are clearly evident through the high volume of interaction on social media, particularly within the comment section of the "Bocor Alus Politik" podcast on YouTube. This phenomenon reflects a significant polarization of public sentiment, which is crucial to map as a basis for evaluating government policy. This study is conducted with the primary objective of developing a sentiment analysis model capable of classifying public opinion regarding the MBG Program into two major categories, namely positive and negative sentiments, utilizing the Naive Bayes algorithm known for its effectiveness in text processing. The research methodology utilizes primary data in the form of collected public comments which undergo a systematic series of text preprocessing stages, including cleaning, tokenization, filtering, and vectorization, to prepare the data for processing. Model performance evaluation is subsequently conducted using the k-fold cross-validation method to ensure the reliability of the classification results. Experimental results indicate that the model successfully achieved an overall accuracy rate of 77.1%. In-depth analysis of per-class performance demonstrates that the model possesses a stronger capability in detecting negative opinions, evidenced by a precision of 0.795, recall of 0.884, and f1-score of 0.837, compared to positive sentiments (precision 0.701, recall 0.545, and f1-score 0.613). Based on these evaluation metrics, it can be concluded that the Naive Bayes algorithm proves to be sufficiently effective in classifying the direction of public sentiment, particularly in identifying public aspirations that are critical of the MBG program.