WITHDRAWN: Performance of bernoulli’s naive bayes classifier in the detection of fake news

Abstract

Abstract Social media has abolished the conventional news sources such as newspaper and TV news channels and became the major source of news. According to Pew Research Centre, which is an Internet research company, 62 percent of adults in the united states gather their news from social media [1] . Social media has become an open platform to spread any type of fake news or propaganda’s all over the world with a single click. Social media platforms like Facebook and Google are using preventive measures to tackle this fake news by using reporting and flagging tools. However, these measures are user-based and only users can report the news as fake or genuine. Machine learning and artificial intelligence may help to create certain powerful algorithms that can detect and remove fake news broadcasting itself. This paper explores a machine learning algorithm named Bernoulli’s Naive Bayes Classifier, which is the extended version of Multinomial Naive Bayes with predictors as Boolean variables i.e. 0 and 1 to detect fake news. Previous studies applied Gaussian Naive Bayes [2] . Our proposed methodology classifies the input data into two classes 0 or 1. ’0′ stands for Fake and ’1′ stands for Genuine news article. Further, it is observed that the results are enhanced as compared to Gaussian Naive Bayes. From the experiments, we observe that the classification results improve by the use of Bernoulli’s Naive Bayes Classifier as compared to Gaussian Naive Bayes. The comparison is done in terms of accuracy, precision, recall, and F1 measure. The accuracy is improved by 10%, precision by 15%, and F1 measure by 6%.