Public Opinion on The MBG Program: Comparative Evaluation of InSet and VADER Lexicon Labeling Using SVM on Platform X

Abstract

This study aims to examine public opinion regarding the MBG program on platform X by utilizing the Support Vector Machine (SVM) algorithm using two sentiment labeling methods, namely InSet Lexicon and VADER Lexicon. The data was then divided into 70% for training and 30% for testing, and extracted using Term Frequency–Inverse Document Frequency (TF-IDF) to convert the text into numerical representations. The SVM model was trained on both labeled data sets to compare their performance based on evaluation metrics such as accuracy, precision, recall, and F1 score. The results show that labeling with VADER produces a more dominant number of neutral sentiments, while InSet Lexicon produces a more balanced distribution between positive, negative, and neutral sentiments. At the modeling stage, SVM with InSet labels achieved an accuracy of 80.10%, with precision of 0.81, recall of 0.80, and an F1 score of 0.79. Meanwhile, SVM with VADER labels achieved an accuracy of 93.83%, precision of 0.94, recall of 0.94, and an F1 score of 0.93. Although VADER showed higher accuracy values, InSet Lexicon is considered more efficient and relevant for sentiment analysis in Indonesia because it is capable of producing more balanced and contextual classifications.