An Approach for Fake News Detection in Social Media Using Hybrid Classifier

Abstract

The users of social networks get ruined by fake news, which brings a vast effect on the offline community. A very significant aim is to improve the trust of data in online social networks by discovering fake news in an appropriately timed manner. A novel technique for identifying fake news on social media has been developed as a result. The feature extraction module receives the preprocessed output after the initial preprocessing stage on the news story. The extracted features are then analyzed using the generative adversarial network (GAN) and neural network (NN) to discover fake scores. Here, the rider shuffled shepherd optimization (RSSO), which was developed by integrating the rider optimization algorithm and the shuffled shepherd optimization algorithm, is used to train the NN classifier. Data labeling is considered an expensive task. GANs are unsupervised, so labeling of data is not required to train them. The score obtained from GAN and NN are utilized for computing the final score, which is evaluated using the weighted average model for detecting fake news. With the highest accuracy of 91.2%, the highest sensitivity of 92.1%, and the highest specificity of 88.8%, the proposed RSSO-based NN + GAN provided improved performance.