SENTIMENT ANALYSIS OF YOUTUBE COMMENTS ABOUT INDONESIAN LGBT USING SUPPORT VECTOR MACHINE AND NAÏVE BAYES ALGORITHMS

Abstract

YouTube is a social media that is widely used by content creators to publish their work, including LGBT content. Because of this content, many viewers end up expressing their opinions through comments. This research aims to see which is the best algorithm between Support Vector Machine using two kernels, linear kernel and RBF kernel or Naive Bayes using multinomial naive bayes seen from confusion matrix. Also, to see which pre-processing is best used for sentiment analysis by dividing pre-processing into several parts. Support Vector Machine using RBF kernel is the best algorithm in this research with 77% accuracy with precision for sentiment -1 74%, recall 72% and f1-score 72%. For sentiment 0, 70% for precision, 81% for recall, and 75% for f1-score. And the last, for sentiment 1, with 90% precision, 77% recall and 83% f1-score. In addition, pre-processing using stemming-tokenizing is the best pre-processing used for sentiment analysis in this research based on the highest average number.