Affiliations: [a] Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Green Fields, Vaddeswaram, Andhra Pradesh, India | [b] Department of Artificial Intelligence and Machine Learning, Saveetha School of Engineering, SIMATS, Chennai 600124, India
Abstract: In recent days, social media is termed a major source for several people residing over the world because of less cost, simpler accessibility, and quick dissemination. However, it comes with dubious trustworthiness and is of high risk in exposing fake news. Hence, the automated discovery of fake news is an essential task. An innovative model is provided to identify fake news considering social media. Here, the BERT model is utilized to perform tokenization in order to produce tokens. Multiple features linked with the data are analyzed for detecting the behavior using the deep model. The features, like Term Frequency-Inverse Document Frequency (TF-IDF), SentiWordNet scores, and sentence level features are obtained to automatically learn the features. Automatic discovery of fake news is done with Aquila Feedback Artificial tree-based Deep Residual Network (AFAT-based DRN). The optimum weight tuning of DRN is executed with AFAT and the AFAT is the fusion of Aquila optimizer (AO) and Feedback artificial tree (FAT). The impact detection of fake news is done with AFAT-based DRN, which helps to detect how many of them shared the fake news. The AFAT-based DRN offered high competence with utmost sensitivity of 92.3%, testing accuracy of 91.6%, and specificity of 91.9%.
Keywords: Fake news, BERT tokenizer, Deep Residual network, Canberra distance, social media