OPINI MASYARAKAT TERHADAP BONUS DEMOGRAFI PADA KANAL YOUTUBE DENGAN METODE TF-IDF, NAÏVE BAYES DAN SMOTE

Abstract

This study examines public opinion on the demographic bonus issue expressed through comments on YouTube channels using the TF-IDF, Naïve Bayes, and SMOTE methods. The data used consists of 870 comments that have been manually labeled into positive and negative sentiments. The research stages include data pre-processing in the form of case folding, removal of non-alphabetic characters, stopword removal, and stemming, then feature extraction using TF-IDF to convert text into numeric representations that can be processed by the algorithm. This study compares the performance of the Naïve Bayes sentiment classification model in two scenarios, namely without and with the application of SMOTE. The SMOTE technique is used to overcome data imbalance between sentiment classes so that the classification results are more balanced and unbiased. The evaluation results show that the model without SMOTE produces an accuracy of 70% but has a very low recall in the positive class. After applying SMOTE, the accuracy increased to 77%, with the highest precision of 0.89 in the negative class and the highest recall of 0.92 in the positive class. The word cloud visualization shows the dominant words that reflect the pattern of public opinion regarding the demographic bonus clearly and informatively. The results of this study can provide a quantitative picture of public perception and be a consideration for policy makers. In the future, this method can be further developed with other algorithms and data from various social media platforms to improve the accuracy and representativeness of sentiment analysis.