A late fusion framework using whale optimization technique and attention-BiLSTM for fake news detection

Recently, fake news detection has garnered interest among the research community. This kind of misconception causes severe political polarization and public mistrust of regulatory agencies in society. Most existing works involved knowledge extraction from Twitter content alone, leading to non-satisfactory performance. This article suggests a unique Late Fusion approach from multiple cues based on Whale Optimization Algorithm-Attention BiLSTM (WOA-ABiLSTM) Networks to obtain high-accuracy performance. The proposed work comprises two stages: Training and testing. During the training and testing phase, we used tweet pre-processing strategies such as stemming, expelling punctuation, stop words, URLs, Twitter controls, and tweet extension. We used Glove word embedding for feature representation and Attention-based Bidirectional LSTM to address challenges associated with extracting meaningful details from tweet data. During the training period, the attention mechanism learns weights on various facets of a phrase when multiple elements are chosen as input. In this work, we late fuse different features of textual data, dissemination, and personal data of posting information to enhance the efficacy of recognizing fake news. Attention weights are learnt using Whale Optimization Algorithm. The proposed WOA-ABiLSTM method demonstrated outstanding performance on the Kaggle dataset with an accuracy of 94%, precision of 90%, recall of 95%, and an F1-Score of 92.5%, while on the FakeNewsNet dataset, it exhibited remarkable results with an accuracy of 92%, precision of 89%, recall of 94%, and an F1-Score of 91.5%. Experimental results unequivocally establish the supremacy of the WOA-ABiLSTM approach, showcasing not only its cutting-edge efficiency in fake news detection but also its substantial advantage over conventional machine learning algorithms, reinforcing its pivotal role in fortifying our information ecosystems against the pervasive threat of misinformation.