Arithmetic Optimized Bi-GRU: A Swift Approach to Combat Fake News in the Digital Sphere

The proliferation of social network services in cyber-physical spaces has made fake news management harder for social service operators. Even with theoretical advances, semantic model processing speed is often a problem. Social media's rapid spread of fake news threatens individuals, communities, and nations, requiring effective detection methods. This study seeks an effective and reliable fake news detection model. We recommend a Bidirectional Gated Recurrent Unit (Bi-GRU) neural network to improve semantic understanding by capturing contextual information from past and future states. We optimize generated tag features with the Arithmetic Optimization Algorithm (AOA) to improve model efficiency and performance. This method uses the distribution behavior of multiplication, division, subtraction, and addition. By selecting relevant features, reducing redundancy, and improving classification performance, the AOA improves the Bi-GRU model. Our results show that the proposed method outperforms previous neural network models in accuracy and processing efficiency, providing a fast and effective solution to fake news. Proposed 70%-30% model. AOA-BiGRU technique accuracy is 0.9412, sensitivity is 0.8889.