Answers are grounded in this paper and cite their sources.
Current traffic flow forecasting is a crucial component within intelligent transportation systems (ITSs). However, most existing studies still cannot be employed on large-scale data flow forecasting in real-world scenarios. In this paper, we design a spatiotemporal neural network called SINN for large-scale traffic flow forecasting tasks in real-world scenarios. First, we propose a transformer-based neural network to integrate the historical traffic flow into further traffic flow. Second, we propose a structural information-based hierarchical graph neural network module suitable for large-scale network structures, which can capture relevant node features within large-scale road networks for forecasting future traffic flow. Finally, we collect a large-scale traffic flow dataset and execute evaluation experiments. The experimental results indicate that SINN is capable of processing datasets with thousands of nodes. Compared with the methods that can execute the same scale datasets, SINN can make significant improvements by 44.29%, 26.54%, and 25.23% in the metrics of MAE, RMSE, and MAPE.
No takes yet. Share an insight, caveat, or question.
Zhu et al. (2024) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: