ANALISIS SENTIMEN KOMENTAR YOUTUBE TENTANG DEMAM BERDARAH DENGUE MENGGUNAKAN NAIVE BAYES

This study aims to analyze public sentiment towards Dengue Hemorrhagic Fever (DHF), a disease that is still a serious health problem in tropical countries such as Indonesia. This problem is explored through sentiment analysis of 1.058 user comments taken from four YouTube videos related to DHF, symptoms, treatment, and recovery. Text preprocessing is applied to the comments, followed by sentiment labeling using InSet Lexicon, and classification using the Multinomial Naive Bayes algorithm. To address class imbalance, the SMOTE (Synthetic Minority Oversampling Technique) method is applied. The dataset is divided into three ratios (70:30, 80:20, and 90:10) to evaluate model performance using Balanced Accuracy, AUC Score, and G-Mean. The result show that the application of SMOTE significantly improves the model’s ability to classify the minority class. The best performance was achieved with a train-test ratio of 70:30, resulting in a Balanced Accuracy of 0.7818, an AUC Score of 0.9357, and a G-Mean of 0.8396. These findings indicate that the combination of Naive Bayes and SMOTE is effective for sentiment classification of imbalanced social media data and can support public health communication strategies