Sentiment Analysis of Indonesian YouTube Comments Using Machine Learning and Deep Learning Algorithms

Abstract

YouTube provides a comment section that reflects public sentiment toward an issue. The large volume of comments makes manual analysis inefficient, requiring automated sentiment analysis. This study compares the performance of machine learning and deep learning algorithms in classifying Indonesian YouTube comments. The dataset consists of 364 comments collected using the YouTube Data API and classified into positive, negative, and neutral sentiments. The data were processed through text preprocessing and represented using TF-IDF for machine learning models, while the deep learning model employed a Long Short-Term Memory (LSTM) architecture. The evaluated algorithms include Naive Bayes, Support Vector Machine (SVM), K-Nearest Neighbor, Decision Tree, Multi-Layer Perceptron (MLP), and LSTM. Model performance was evaluated using accuracy, precision, recall, and F1-score. The results show that SVM achieved the best performance, followed by MLP and Naive Bayes, while LSTM performed the worst due to limited data size and class distribution. This study indicates that traditional machine learning remains effective for sentiment analysis on small Indonesian-language datasets.