Automatic Fake News Detection on Social Networks using Softsign Activation based Long-Short Term Memory

In past few years, social media have shortened a technique of data transformation like news related to traditional processes. A multimodal method is proposed in this paper for the detection of fake news and classifying the news into actual or forged. The dataset ISOT and Buzzfeed is employed and data is preprocessed by several methods. The video/image, audio and text features are extracted through separate techniques which is concatenated and provided to classification Softsign Activation-Long-Short Term Memory (SA-LSTM) for classification process. SA-LSTM performance is valued with accuracy, precision, sensitivity, specificity and f1-score. Proposed technique accomplished higher accuracy of 99.64% and 98.55% for ISOT and Buzzfeed dataset respectively that is higher than existing methods like LSTM-Levy Flight (LSTM-LF), Multi Support Vector Machine (MSVM) and Random Forest (RF).