Hoax News Detection Analysis using IndoBERT Deep Learning Methodology

Today's massive interaction on the social media platform results in a lot of information circulating, making the news abundant and easily accessible. We also witness much contradictory news fabricated according to the content creator agenda. This phenomenon makes it more likely that people get hoax news. The public is asked to be more vigilant and not easily trust the circulated information. The existence of hoax news can damage a company's reputation that has long been built. This research aims to detect hoax news using IndoBERT, Support Vector Machine, and Naïve Bayes by classifying news in Indonesian and finding the best model. The collected data includes various topics on news sites such as Detik, Liputan 6, Kompas, Cek Fakta, and Turnbackhoax. A total of 2000 news contained 1000 hoax news data and 1000 non-hoax news data. The best accuracy is the IndoBERT algorithm, with a value of 90%. We analyze this result and find that IndoBERT had the best performance detecting hoaxes by detecting contextual aspects related to more complex patterns.