Perbandingan Algoritma Naïve Bayes, Support Vector Machine, dan Random Forest Untuk Analisis Sentimen Komentar Politik Youtube

Santoso, Bayu Aji
Nugroho, Bangkit Indarmawan
Asyfiya, Dzurrotu Umi

TIN: TERAPAN INFORMATIKA NUSANTARA
Vol 6 No 4 (2025): September 2025

Sentiment analysis is an important field in natural language processing that is widely used to understand public opinion on social media. This study compares the performance of three machine learning algorithms, namely Naïve Bayes, Support Vector Machine (SVM), and Random Forest, in analyzing YouTube comment sentiment. The dataset consists of 15,257 comments obtained from the Indonesia Lawyers Club (ILC) and Rakyat Bersuara channels. The research process includes preprocessing stages (cleaning, case folding, tokenizing, normalization with a slang dictionary, stopword removal, and stemming), data labeling with a Lexicon-based approach using InSet Lexicon, data division with a ratio of 80% training data and 20% test data, and evaluation using accuracy, precision, recall, and F1-score metrics, complemented by K-fold cross validation tests. The results of the sentiment distribution show a dominance of negative sentiment at 43.2%, followed by neutral at 34.9%, and positive at 21.9%. Model evaluation showed that SVM excelled with 83.52% accuracy, 83.55% precision, 83.52% recall, and 83.52% F1-score, followed by Random Forest with 77.20% accuracy, while Naïve Bayes achieved the lowest result at 64.71%. The K-Fold test further strengthened these results, with the best accuracy of 84.14% for SVM. Thus, SVM can be concluded as the most effective algorithm for analyzing political comment sentiment on YouTube.