Fake News Detection System using Machine Learning with Recurrent Neural Networks

Abstract

The rapid growth of fake news on digital platforms has posed a significant threat to information integrity and societal trust. A number of machine learning techniques were proposed and tested in this study include Naïve Bayes, Long Short-Term Memory (LSTM) and Bi Directional RNNs for detecting fake news. The main aim is to assess and compare the power of the models in detecting deceptive content. The primary objective is to evaluate and compare the effectiveness of these models in accurately identifying deceptive content. The Naïve Bayes model, serving as a baseline, achieved an accuracy of 93.42%, demonstrating its utility in handling simple text classification tasks. However, the introduction of deep learning techniques significantly enhanced detection performance. The LSTM model, designed to capture long-term dependencies in text data, achieved an accuracy of 99.62%. The GRU model, a simplified variant of LSTM with comparable performance, achieved a slightly higher accuracy of 99.83%. The Bi-directional RNN, which processes input data in both forward and backward directions, outperformed all other models, and recorded an accuracy of 99.85%. With the maximum accuracy provided by the Bi-directional RNN model, these results demonstrated the capability of deep learning models in detecting fake news.  Findings from this research will be useful in the fight against cybersecurity threats and in limiting the dissemination of false information.