FIND: Fake Information and News Detections using Deep Learning

Fake news detection is very difficult while its spread is simple and has vast repercussions. To tackle this problem we propose a model which detects fake information and news with the help of Deep Learning and Natural Language Processing. A Deep Neural Network on self scraped data set is trained and by using Natural Language Processing the correlation of words in respective documents is found and these correlations serve as initial weights for the deep neural network which predicts a binary label to detect whether the news is fake or not. In this work we have successfully used Recurrent Neural Network and Long Short-Term Memories and Grated Recurrent Units to test for classification. Tensorboard is used for implementation of the proposed framework and provide visualizations for the neural network. Confusion matrix and classification reports show that accuracy score of upto 94 percent can be achieved using LSTM model. The tradeoff is the increased time requirement. Since the fake news can be established based on the learning model, a good training set is mandatory. The results show that the proposed framework is able to adequately present accurate result.