Fake News Detection Using Optimized Deep Learning Model Through Effective Feature Extraction

Internet plays a vital role in the propagation of data through public networks or websites due to which Fake news is produced for commercial and political benefits to misinform and fascinate readers. In order to stop the scattering of fake news, detection techniques are researched based on deep learning in natural language processing tasks. The deep learning model proposed in this study combines LSTM and Bi-directional LSTM with one-hot encoding representation for classification of fake news. The model is effectively authenticated on the political news articles dataset consisting of 240 websites (URLs), body, headline, and label. TensorFlow framework utilizes built-in Keras deep learning libraries that is having an online community on TensorFlow Kaggle repository. The outcome demonstrates the Bi-directional LSTM model's advantage over the Unidirectional LSTM model in terms of accuracy