Multi Class News Accuracy Assessment VIA LSTM

Abstract

For information about a public event, media play an important role. These days the internet is a major source of information dispersal on a huge scale. The authenticity of this data is a question for news agencies and business tycoons. The credibility of news is important for companies to make decisions regarding their policies and decisions. Fake news and information are generally spread by industrial, educational, and political parties to mislead the user. Automated credibility analysis of information articles is a hot research topic in this era of social media. Deep learning models are broadly used for linguistic modelling. Typical deep learning models along with recurrent neural networks (RNN) and Convolutional neural networks (CNN) can hit upon complicated textual data patterns. Long Short-Term Memory (LSTM) is considered as a sequence prediction-dependent Recurrent Neural Networks (RNN) used to research variable-duration sequential records. The proposed method gives a fake news detection model version based on LSTM-Recurrent Neural Network. The result indicates that the proposed method shows 99% accuracy in fake news multiclass classification.