Reliable social media framework: fake news detection using modified feature attention based CNN-BiLSTM

The social media platforms leads to proliferation of fake news and spreading of these false information creates high negative impact on the society. To overcome these consequences, it is essential to develop automatic detection of fake news in order to protect the environment. When compared with traditional Machine Learning (ML) techniques, Deep Learning (DL) algorithms showed an encouraging outcomes in Natural Language Processing (NLP). So, the proposed system implements Modified Feature Attention based Convolutional Neural Network-Bidirectional Long Short Term Memory (CNN-BiLSTM) approach for fake news classification. By using these datasets, the fake and real news are classified by using CNN-BiLSTM model along with attention mechanism. CNN already has good learninig skills, when combined with BILSTM, it gets really improved in efficiency and precision. Additionally, modified feature attention mechanism is involved, to concentrate and extract on specific information by using feature and correlation matrix. Further, the efficacy of the proposed model is identified by using performance metrics such as precision, recall, F1-score and accuracy. In order to predict the efficiency of the proposed system, it is compared with other conventional algorithms.