An Efficient Fake News Identification System Using A-SQUARE CNN Algorithm

Nowadays, one amongst the Semantic Web (SW), called the Social Media (SM), renders a platform for quick and also faultless access to information. Millions of individuals considered it as the chief news source. With the escalating popularity of SM and web-based forums, the distribution of Fake News (FN) has grown enormously in current times, which become a chief threat to an assortment of sectors and agencies. Numerous solutions and techniques were commenced for identifying FN quickly enough to end the spreading of FN. Nevertheless, they all failed on account of a lack of classification accuracy and high training time. To trounce these issues, this paper proposes a FastText Modified Embeddings from Language Models (FT-MELMo) based efficient FN identification system (FNIS). Specifically, the proposed system learns the semantic representation of data by utilizing FT-MELMo embedding. FT and MELMo utilize character n-grams and Bi-GRU, correspondingly to maximize the classification accuracies and also minimize the training time. Lastly, a higher-dimensional embedding feature illustration is inputted to the AMSGradAdamNC based Convolutional Neural Network (A²CNN) classifier, which classifies the news as FN or real news effectively. The suggested A²CNN classifier achieved a classification accuracy of 96.54 percent with a low FDR of 4.32 percent. When compared to existing classifiers, the suggested A²CNN exhibits a low FPP and FNR. On the basis of the performance comparison, the suggested FNIS is found to be extremely accurate, robust, and faster than other classifiers.