Comparative analysis of text classification using naive bayes and support vector machine in detecting negative content in Indonesian twitter

Abstract

Research on the detection of social media content, especially Twitter, has been done. Twitter content detection is based on classifying content or words made by users (tweets) into two groups, namely positive and negative. Research to detect negative content or harsh words in Indonesian tweets is still rare. There are several studies that have been conducted, the detection of negative words is only limited to certain categories, such as pornography, hate speech, and others, so that if the negative word only includes one category, then if there are other negative words that do not belong to that category, this word will not be detected. This is a challenge for researchers to classify texts in Indonesian. In some research, to be able to detect and separate negative words and positive words in Indonesian, the Naive Bayes (NB) and Support Vector Machine (SVM) proved to produce better performance among other algorithms. Therefore, this paper aims to analyze the comparison of the results that have been achieved about detecting negative content on Indonesian twitter using NB and SVM. First, this paper will briefly explain the NB and SVM, then proceed with an explanation of the general research framework that has been carried out. In the results and discussion section, a comparison of the results achieved by existing researchers is explained. And based on these results, another approach will be proposed to detect negative content on Indonesian twitter.