Sentiment Analysis and Topic Modelling of Bjorka Using Support Vector Machine and Latent Dirichlet Allocation
Muhammad Muhajir (1,3*), Dedi Rosadi(2)

(1*) Department Mathematics, Gadjah Mada University, Sendowo, Sinduadi, Kec. Mlati, Kabupaten Sleman, Yogyakarta 55281, Indonesia.
*muhammadmuhajir[at]mail.ugm.ac.id
(2) Department Mathematics, Gadjah Mada University, Sendowo, Sinduadi, Kec. Mlati, Kabupaten Sleman, Yogyakarta 55281, Indonesia.
(3)Department Statistics, Islamic University of Indonesia, Jl. Teknika, Krawitan, Umbulmartani, Kec. Ngemplak, Kabupaten Sleman,Yogyakarta 55584, Indonesia

Abstract

A wide range of data is now easily accessible via the microblogging service Twitter thanks to the rapid advancement of technology. The bjorka controversy, one of the most talked-about topics right now, has generated numerous comments from the general public and thus has risen to the top. The Bjorka phenomenon is an obvious example of cybercrime, with a sharp uptick in incidents occurring in Indonesia during the Covid-19 pandemic. Sentiment analysis employing the Support Vector Machine technique allows for the statistical analysis of public opinion about Bjorka as it appears on the Twitter social network. Latent Dirichlet Allocation (LDA) will be used to analyze the sentiment analysis with SVM results, which have been separated into positive and negative sentiments. In this study, using LDA for sentiment analysis resulted in an accuracy of 89.5%. Dismantling government data, including personal data and government crimes, was the most positively predicted topic, with 75.2% of all predictions leaning in that direction. It is hoped that the government will be able to use the information gleaned from this study to better understand the public^s perspective and the trust deficits that need to be addressed.

Keywords: Bjorka, Cybercrime, SVM, LDA

Topic: MATHEMATICS AND STATISTICS