Klasifikasi Komentar Toxic Sara pada Tweet Bahasa Indonesia Menggunakan Adaptive Boosting (Adaboost)

Abstract

Social media is currently growing rapidly and has become part of communication. Social media provides its own space for its users to be free of expression, especially in providing comments on what is seen and felt in written form. However, because of this freedom in commenting, social media has become a place filled with toxic comments. Classify toxic comments It is very important to protect social media users. However, the process of identifying toxic comments certainly takes a long time if done manually. Therefore, an approach is needed that can automate the process of classifying toxic comments. This research aims to classify SARA toxic comments in Indonesian using fastText feature extraction and the adaptive boosting method (Adaboost). The dataset used in this study is 1000 data with 800 training data and 200 testing data sourced from Twitter social media which consists of four classes namely ethnicity, religion, race and intergroup. Based on testing, this research produces an accuracy of 91.5%. From these results it can be concluded that the system created using the Adaboost method and the fastText feature extraction is good enough at classifying SARA toxic comments.