FAKE NEWS DETECTION USING MACHINE LEARNING

Abstract

The problem of Fake news has evolved much faster in the recent years. Social media has dramatically changed its reach and impact as a whole. On one hand, it’s low cost, and easy accessibility with rapid share of information draws more attention of people to read news from it. On the other hand, it enables wide spread of Fake news, which are nothing but false information to mislead people. As a result, automating Fake news detection has become crucial in order to maintain robust online and social media. Artificial Intelligence and Machine learning are the recent technologies to recognize and eliminate the Fake news with the help of Algorithms.

In this work, Machine-learning methods are employed to detect the credibility of news based on the text content and responses given by users. A comparison is made to show that the latter is more reliable and effective in terms of determining all kinds of news. The method applied in this work is highest posterior probability of tokens in the response of two classes. It uses frequency-based features to train the Algorithms including Support Vector Machine, Passive Aggressive Classifier, Multinomial Naïve Bayes, Logistic Regression and Stochastic Gradient Classifier. This work also highlights a wide-range of features established recently in this area that gives a clearer picture for the automation of this problem. I have conducted an experiment in this work to match the lists of Fake related words in the text of responses, to find out whether the response based detection is a good measure to determine the credibility or not. The results were found to be very promising and have v

scope for more research in the area. Linear SVM and Stochastic Gradient Classifier algorithm with Tf-Idf vector achieved Accuracy and ROC Area under curve above 90% and 95% respectively. This work can be used as a significant building block for determining the veracity of Fake news