Analisis Sentimen Program Makan Bergizi Gratis (MBG) untuk Anak Sekolah Menggunakan Algoritma Support Vector Machine

Abstract

The Support Vector Machine with a linear kernel has proven to be highly effective and balanced in classifying public perceptions of the Free Nutritious Meals Programme (MBG). This conclusion was drawn from an analysis of public opinion on the social media platform X, which aimed to map public sentiment towards the programme. The analysis was conducted using a linear kernel SVM algorithm with the SEMMA methodology, utilising 1,000 initial data points from X, which were filtered down to 578 binary data points (401 positive and 177 negative) through a TF-IDF-based cleaning and labelling process. The data was split in an 80:20 ratio and tested across five variations of the hyperparameter C. The best configuration was achieved at C=1, with a global accuracy of 96.55 per cent, a negative F1-score of 94.44 per cent, and a positive F1-score of 97.50 per cent. In terms of content, positive sentiment centred on improvements in children’s nutrition (keywords ‘nutrition’ and ‘food’), whilst negative sentiment related to the risk of food poisoning and budget efficiency (keywords ‘poison’ and ‘budget’). Given this high and balanced performance across both positive and negative classes, the linear kernel SVM is considered reliable for the task of classifying public policy perceptions.