Identifying Fake News with Various Machine Learning Model

Abstract

News is basically information, information regarding recent or current events. There are various platforms or sources of news like: print media, TV channels, digital media, social media and films. It is obvious that news floating in these platforms are not always correct. Sometimes these platforms contain false or incorrect information, which is termed as fake news. Fake news can be in the form of images, videos, audios and texts. Generally fake news is generated due to a machine or human intervention. In order to identify fake news in various platforms this paper reviewed several techniques based on a machine learning and artificial intelligence. This paper is going to review various machine learning based approaches and some of the models are Naïve Bayes Classifier that has been tested in a software system with a data set of Facebook news posts and had achieved an accuracy of 74 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> , Convolutional Neural Network (CNN) for image visualization scored a mean accuracy of 92.85%, Recurrent Neural Network (RNN) for audio and text visualization had achieved an accuracy of 93 % on the datasets collected from Brussels terrorist attack in 2016. Some of the hybrid models are like for the early detection of false news on twitter using propagation path classification using combined CNN and RNN model had achieved an accuracy of 85%, Meta Optimization Semantic Evolutionary Search model (MOSES) scored a mean accuracy of 63%, Capsule Neural Network (CapsNet) model for an image, audio, video, text mining scored overall accuracy of 99.8%. Total twelve machine learning models have been represented with proper data visualization of their accuracy percentage and found that Artificial Neural Network based Capsule Neural Network (CapsNet) model is the best with a highest accuracy of 99.8% over LIAR and ISOT set of datasets.