Fake News Detection using Naive Bayes

Abstract

The issue of fake news arises every year. Moreover, the enhancement and evolution of technologies enable the news to be manipulated by irresponsible people. However, it is not deniable that somehow this technology impacts our daily life. Nowadays, people get the latest news through the social media platforms as it is free, easy to access, and fast. However, not all the news on social media is reliable, and some fake news are spread to mislead the readers. Fake news can disseminate information to confuse people to believe things that are not true. In Natural Language Processing, text processing such as regular expression, removing the stop words and lemmatization are done before the data is being transformed into N-grams using TF-IDF and Count Vectorizer. Therefore, this paper aimed to review the fake news detection using the Naive Bayes algorithms. Results shows that Naive Bayes with n-gram gives a slight increase in the accuracy of TF-IDF and Count Vectorizer. It proves that TF-IDF Vectorizer can detect fake news better as it has higher precision of 94 % whereas Count Vectorizer can detect both fake news and real news in quite a balance.