Period Extraction for Traffic Flow Prediction
Q Wang, C Chen, L Zhang, X Song, H Li… - … on Algorithms and …, 2023 - Springer
Q Wang, C Chen, L Zhang, X Song, H Li, Q Zhao, B Niu, J Gu
International Conference on Algorithms and Architectures for Parallel Processing, 2023•SpringerDue to the particularity of “Tourist chartered Buses, Liner Buses and Dangerous Goods
Transport Vehicles”(“TLD Vehicles”), traffic accidents will bring serious losses. Therefore,
traffic flow prediction for “TLD Vehicles” has become an urgent need for traffic management
departments. Different from the ordinary traffic flow prediction problem, the traffic flow for
“TLD Vehicles” has the characteristics of sparsity in the spatial dimension. The ordinary
spatial feature extraction method will capture useless node information and affect the …
Transport Vehicles”(“TLD Vehicles”), traffic accidents will bring serious losses. Therefore,
traffic flow prediction for “TLD Vehicles” has become an urgent need for traffic management
departments. Different from the ordinary traffic flow prediction problem, the traffic flow for
“TLD Vehicles” has the characteristics of sparsity in the spatial dimension. The ordinary
spatial feature extraction method will capture useless node information and affect the …
Abstract
Due to the particularity of “Tourist chartered Buses, Liner Buses and Dangerous Goods Transport Vehicles” (“TLD Vehicles”), traffic accidents will bring serious losses. Therefore, traffic flow prediction for “TLD Vehicles” has become an urgent need for traffic management departments. Different from the ordinary traffic flow prediction problem, the traffic flow for “TLD Vehicles” has the characteristics of sparsity in the spatial dimension. The ordinary spatial feature extraction method will capture useless node information and affect the prediction accuracy. The traffic data of “TLD Vehicles” has significant periodic characteristics in the time dimension. Most of the traditional traffic prediction methods extract time characteristics through artificially set cycles, which has certain limitations. In this paper, a Period Extraction model for Traffic Flow Prediction is proposed to solve the problem of data sparsity and insufficient periodic feature extraction ability of traffic flow prediction. The model uses sparse graph convolution combined with Transformer to extract spatial features, uses sequence decomposition and auto-correlation attention mechanism to extract temporal features, and obtains prediction results through stacked spatio-temporal modules. The experimental results show that the proposed algorithm can extract the traffic f low data feature information with stable characteristics more effectively and improve the accuracy of the network model for traffic prediction.
Springer
Showing the best result for this search. See all results