Detection of Twitter Fake News using Efficient Soft-Capsule and Improved BiGRU Architecture

Abstract

Social media platforms, such as Twitter, are vulnerable to the spread of fake news, which can have significant consequences on people's daily lives. To combat this issue, various techniques have been developed to detect fake news, but they often have limitations, including low performance and high training times. To overcome these limitations, a new enhanced fake news detection technique is proposed, which utilizes an efficient soft-capsule and improved BiGRU model. This technique combines image and text data from the Twitter Fake News Detection (2ter-Fk-Nus) Model dataset, processing each modality separately with different pre-processing and feature extraction techniques. The extracted features are then optimized using the Binary Guided Whale–Dipper Throated Optimizer (BGW-DTO) method, and finally, the features from both text and image are fused using Cross-model Fusion (CmF) to predict whether a tweet is fake or real. The proposed model, Improved BiGRU efficient soft-capsule 2ter-Fk-Nus(IBiG-EcnTSCaps 2ter-Fk-Nusd), achieves an overall accuracy of 99.95%, outperforming other related techniques.