A novel hybrid neural network for high-accuracy vehicle-to-infrastructure network traffic prediction

Executive Summary

The article discusses a novel hybrid neural network model designed for high-accuracy vehicle-to-infrastructure (V2I) network traffic prediction. It focuses on transforming vehicle trajectory data into knowledge-driven V2I traffic datasets using path loss and speed loss models. The integration of GhostNet modules allows for efficient feature extraction and scalable spatial analysis. The model, named gCNN-BiLSTM-MHA, combines gCNN, BiLSTM, and multi-head attention mechanisms to capture spatial features, bidirectional temporal dependencies, and key data interactions. Extensive experiments on multiple benchmark datasets demonstrate that this model consistently outperforms existing methods in terms of accuracy, efficiency, and stability. The research advances AI methodologies for knowledge-intensive V2I communication system research, validating the model's scalability and reliability across various datasets.