Abstract:Traffic congestion event prediction is an important yet challenging task in intelligent transportation systems. Many existing works about traffic prediction integrate various temporal encoders and graph convolution networks (GCNs), called spatio-temporal graph-based neural networks, which focus on predicting dense variables such as flow, speed and demand in time snapshots, but they can hardly forecast the traffic congestion events that are sparsely distributed on the continuous time axis. In recent years, neural point process (NPP) has emerged as an appropriate framework for event prediction in continuous time scenarios. However, most conventional works about NPP cannot model the complex spatio-temporal dependencies and congestion evolution patterns. To address these limitations, we propose a spatio-temporal graph neural point process framework, named STGNPP for traffic congestion event prediction. Specifically, we first design the spatio-temporal graph learning module to fully capture the long-range spatio-temporal dependencies from the historical traffic state data along with the road network. The extracted spatio-temporal hidden representation and congestion event information are then fed into a continuous gated recurrent unit to model the congestion evolution patterns. In particular, to fully exploit the periodic information, we also improve the intensity function calculation of the point process with a periodic gated mechanism. Finally, our model simultaneously predicts the occurrence time and duration of the next congestion. Extensive experiments on two real-world datasets demonstrate that our method achieves superior performance in comparison to existing state-of-the-art approaches.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the important but challenging task of traffic congestion event prediction in intelligent transportation systems. Most of the existing traffic prediction works use Spatio - Temporal Graph - based Neural Networks. These methods mainly focus on predicting dense variables in time snapshots, such as flow, speed and demand, but perform poorly in predicting traffic congestion events sparsely distributed on the continuous time axis. In addition, although the Neural Point Process (NPP) performs well in event prediction in continuous - time scenarios, most traditional NPP methods have difficulty in modeling complex spatio - temporal dependencies and congestion evolution patterns. To solve these problems, the authors propose a new spatio - temporal graph neural point process framework, called STGNPP (Spatio - Temporal Graph Neural Point Process), which is specifically used for traffic congestion event prediction. Specifically, STGNPP improves the existing methods in the following ways: 1. **Spatio - temporal graph learning module**: A spatio - temporal graph learning module is designed to fully capture spatio - temporal dependencies over a long - time range from historical traffic state data. 2. **Continuous Gated Recurrent Unit**: The extracted spatio - temporal latent representation and congestion event information are input into a Continuous Gated Recurrent Unit (CGRU) to model the congestion evolution pattern. 3. **Periodic gate mechanism**: In order to make full use of periodic information, the calculation of the point process intensity function is improved and a periodic gate mechanism is introduced. 4. **Simultaneous prediction of occurrence time and duration**: Finally, the model can simultaneously predict the occurrence time and duration of the next congestion event. Through extensive experiments on two real - world datasets, the authors prove that the proposed STGNPP model significantly outperforms the existing state - of - the - art methods in prediction performance.

Spatio-Temporal Graph Neural Point Process for Traffic Congestion Event Prediction

A Spatial-Temporal Gated Hypergraph Convolution Network for Traffic Prediction

Periodic Shift and Event-aware Spatio-Temporal Graph Convolutional Network for Traffic Congestion Prediction

Spatio-Temporal Graph Neural Network for Traffic Prediction Based on Adaptive Neighborhood Selection

Spatial-temporal graph neural network based on gated convolution and topological attention for traffic flow prediction

A dynamical spatial-temporal graph neural network for traffic demand prediction

STGMN: A gated multi-graph convolutional network framework for traffic flow prediction

Spatio-Temporal Pivotal Graph Neural Networks for Traffic Flow Forecasting

Adaptive Spatio-temporal Graph Neural Network for traffic forecasting

Spatial–Temporal Complex Graph Convolution Network for Traffic Flow Prediction

Make More Connections: Urban Traffic Flow Forecasting with Spatiotemporal Adaptive Gated Graph Convolution Network

Spatial-Temporal Interactive Graph Neural Network for Traffic Forecasting.

Spatio-Temporal Graph-TCN Neural Network for Traffic Flow Prediction

Traffic Flow Prediction Via Spatial-Temporal Geometric Graph Convolutional Network

Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting

A Spatio-Temporal Graph Neural Network Approach for Traffic Flow Prediction

Spatial–Temporal-Correlation-Constrained Dynamic Graph Convolutional Network for Traffic Flow Forecasting

A Spatiotemporal Graph Neural Network with Graph Adaptive and Attention Mechanisms for Traffic Flow Prediction

Dual Dynamic Spatial-Temporal Graph Convolution Network for Traffic Prediction

Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecasting

Global-Aware Enhanced Spatial-Temporal Graph Recurrent Networks: A New Framework For Traffic Flow Prediction