PERBANDINGAN KINERJA SVM, KNN, DAN NAÏVE BAYES PADA ANALISIS SENTIMEN TWEET MBG

Abstract

This study explores the effectiveness of three machine learning classifiers—Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Naïve Bayes—in analyzing public sentiment toward the Free Nutritious Meal Program using data from the X (Twitter) platform. Tweets were collected via Tweet Harvester by applying keyword-based filtering over the August–October 2025 period. Prior to model implementation, the textual dataset underwent comprehensive preprocessing, including data cleaning, case normalization, lexical standardization, tokenization, stopword elimination, and stemming. Sentiment labels were generated using a lexicon-based approach to distinguish between positive and negative opinions. The processed data were divided into training and testing subsets for classification. Model performance was evaluated using accuracy metrics derived from the confusion matrix. The results show that SVM outperformed the other models with an accuracy of 91.7%, followed by Naïve Bayes at 79.6% and KNN at 79.3%, indicating the strong capability of SVM in handling complex textual representations in social media sentiment analysis.