Abstract:Traffic flow forecasting is a fundamental research issue for transportation planning and management, which serves as a canonical and typical example of spatial-temporal predictions. In recent years, Graph Neural Networks (GNNs) and Recurrent Neural Networks (RNNs) have achieved great success in capturing spatial-temporal correlations for traffic flow forecasting. Yet, two non-ignorable issues haven't been well solved: 1) The message passing in GNNs is immediate, while in reality the spatial message interactions among neighboring nodes can be delayed. The change of traffic flow at one node will take several minutes, i.e., time delay, to influence its connected neighbors. 2) Traffic conditions undergo continuous changes. The prediction frequency for traffic flow forecasting may vary based on specific scenario requirements. Most existing discretized models require retraining for each prediction horizon, restricting their applicability. To tackle the above issues, we propose a neural Spatial-Temporal Delay Differential Equation model, namely STDDE. It includes both delay effects and continuity into a unified delay differential equation framework, which explicitly models the time delay in spatial information propagation. Furthermore, theoretical proofs are provided to show its stability. Then we design a learnable traffic-graph time-delay estimator, which utilizes the continuity of the hidden states to achieve the gradient backward process. Finally, we propose a continuous output module, allowing us to accurately predict traffic flow at various frequencies, which provides more flexibility and adaptability to different scenarios. Extensive experiments show the superiority of the proposed STDDE along with competitive computational efficiency.

A Continuous Encoder-Decoder Method for Spatial-Temporal Forecasting

STONet: A Neural-Operator-Driven Spatio-temporal Network

Spatial-Temporal Graph Discriminant AutoEncoder for Traffic Congestion Forecasting

Continuous Spatial-Temporal Convolutional Networks Based Neural Control Differential Equations for Traffic Forecasting

Continuously Evolving Graph Neural Controlled Differential Equations for Traffic Forecasting

CAG-NSPDE: Continuous Adaptive Graph Neural Stochastic Partial Differential Equations for Traffic Flow Forecasting

Multi-View Neural Differential Equations for Continuous-Time Stream Data in Long-Term Traffic Forecasting

Unveiling Delay Effects in Traffic Forecasting: A Perspective from Spatial-Temporal Delay Differential Equations

Multi-Branch Spatial-Temporal Decoupling Neural Network for Traffic Forecasting.

Spatiotemporal Dynamic Multi-Hop Network for Traffic Flow Forecasting

LSTGCN: Inductive Spatial Temporal Imputation Using Long Short Term Dependencies

Effective improvement of multi-step-ahead flood forecasting accuracy through encoder-decoder with an exogenous input structure

Multi-Scale Spatio-Temporal Attention Networks for Network-Scale Traffic Learning and Forecasting

Location and time embedded feature representation for spatiotemporal traffic prediction

A Novel Artificial Neural Network for Flood Forecasting Based on Deep Learning Encoder-Decoder Architecture

Link prediction by continuous spatiotemporal representation via neural differential equations

Spatial-Temporal Hypergraph Neural Network for Traffic Forecasting

STEGNN: Spatial-Temporal Embedding Graph Neural Networks for Road Network Forecasting

Dynamic Spatial-Temporal Embedding Via Neural Conditional Random Field for Multivariate Time Series Forecasting

Forecast Network-Wide Traffic States for Multiple Steps Ahead: A Deep Learning Approach Considering Dynamic Non-Local Spatial Correlation and Non-Stationary Temporal Dependency

CRFST-GCN: A Deeplearning Spatial-Temporal Frame to Predict Traffic Flow