ChebNet Traffic Flow Prediction Model Based on Non-local Spatio-temporal Correlation Matrix

In traffic flow prediction tasks, accurately identifying spatio-temporal dependencies with strong correlations to target nodes constitutes a fundamental research challenge. Existing studies reveal that graph convolutional neural networks (GCNs) constrained by conventional adjacency matrices demonstrate limited capabilities in capturing comprehensive spatio-temporal dependencies, thereby compromising prediction accuracy. To address these limitations, a ChebNet model based on a non-local spatio-temporal correlation matrix (C-ChebNet) is proposed for traffic flow prediction. The technical contributions are threefold: First, we introduce temporal and spatial delay parameters to establish a novel Spatio-temporal Cross-correlation Function (ST-CCF) index for quantifying the correlation between nodes. Furthermore, ST-CCF is employed to construct non-local spatio-temporal adjacency matrices, effectively replacing conventional Euclidean distance-based matrices in ChebNet architecture. Finally, public datasets PeMS04 and PeMS07 are selected to evaluate the model's performance. Experimental results demonstrate that the proposed model achieves the best prediction accuracy with lower time complexity compared to other models such as Graph Convolutional Networks (GCN), Graph Attention Networks (GAT), Spatio-temporal Graph Convolutional Networks (STGCN) and baseline ChebNet model.