Sentiment Analysis: Classifying Public Comments on YouTube in Disaster Management Simulation in Indonesia Using Naïve Bayes and Support Vector Machine

Abstract

The objective of this research is to classify public comments on YouTube related to disaster preparedness simulations through sentiment analysis. The research process included data collection, labeling, pre-processing, and classification. Support Vector Machine (SVM) and Naïve Bayes algorithms were used for classification. Following manual labeling of 204 datasets, the breakdown of sentiment was as follows: 112 positive, 43 negative, and 49 neutral. The evaluation involved two scenarios: performance testing and sensitivity testing. Performance testing, conducted on pre-processed datasets, revealed that Naïve Bayes Classifier (NBC) achieved an accuracy rate of 80.4%, with the best execution time of 0.0097 seconds. In contrast, the Support Vector Machine (SVM) achieved the highest accuracy rate of 72.3%, albeit with a longer worst-case execution time of 193.48 seconds. Furthermore, in the results of sensitivity measurements using the dataset without going through the preprocessing stages, each method was able to show the best results with a value of 100%.