Abstract:The complex spatial-temporal correlations in transportation networks make the traffic forecasting problem challenging. Since transportation system inherently possesses graph structures, many research efforts have been put with graph neural networks. Recently, constructing adaptive graphs to the data has shown promising results over the models relying on a single static graph structure. However, the graph adaptations are applied during the training phases and do not reflect the data used during the testing phases. Such shortcomings can be problematic especially in traffic forecasting since the traffic data often suffer from unexpected changes and irregularities in the time series. In this study, we propose a novel traffic forecasting framework called Progressive Graph Convolutional Network (PGCN). PGCN constructs a set of graphs by progressively adapting to online input data during the training and testing phases. Specifically, we implemented the model to construct progressive adjacency matrices by learning trend similarities among graph nodes. Then, the model is combined with the dilated causal convolution and gated activation unit to extract temporal features. With residual and skip connections, PGCN performs the traffic prediction. When applied to seven real-world traffic datasets of diverse geometric nature, the proposed model achieves state-of-the-art performance with consistency in all datasets. We conclude that the ability of PGCN to progressively adapt to input data enables the model to generalize in different study sites with robustness.

What problem does this paper attempt to address?

This paper attempts to solve the problem of complex spatio - temporal correlations in traffic prediction. Specifically, the inherent graph - structure characteristics of traffic networks make traffic prediction a challenging problem. Many existing studies have already utilized Graph Neural Networks (GNNs) to deal with this problem, and some recent methods have achieved remarkable results by constructing graph structures adapted to the data. However, the adaptive graph structures constructed in the training phase of these methods do not reflect the data characteristics in the testing phase, which is especially important in traffic prediction because traffic data often has unexpected changes and irregularities in time series. To solve the above problems, the authors propose a new traffic prediction framework - Progressive Graph Convolutional Network (PGCN). The main innovation of PGCN lies in its ability to dynamically construct graph structures according to online input data during both the training and testing phases. Specifically, PGCN constructs a progressive adjacency matrix by learning the trend similarity between nodes and combines dilated causal convolution and gated activation unit to extract temporal features. In addition, through residual connections and skip connections, PGCN can effectively perform traffic prediction. The main contributions of the paper include: 1. **Proposing a new spatio - temporal traffic prediction model**: PGCN, whose main components are the progressive graph convolution module and dilated causal convolution. The progressive graph convolution module extracts spatial features by capturing spatio - temporal correlations that change over time, while dilated causal convolution is used to extract temporal features. 2. **Constructing progressive graphs**: These graphs capture the trend similarity between node signals through parameterized cosine similarity and update edges and weights according to online input data at each time step. 3. **Experimental verification**: Experiments were carried out on seven real - world traffic datasets, and the results show that the model can achieve state - of - the - art performance on all datasets and has strong generalization ability and robustness at different research locations. In conclusion, this paper aims to solve the problem of spatio - temporal correlations that change over time in traffic prediction by introducing the Progressive Graph Convolutional Network, thereby improving the accuracy and robustness of prediction.

PGCN: Progressive Graph Convolutional Networks for Spatial-Temporal Traffic Forecasting

PGCN: Progressive Graph Convolutional Networks for Spatial–Temporal Traffic Forecasting

A Spatial-Temporal Gated Hypergraph Convolution Network for Traffic Prediction

Probabilistic spatio-temporal graph convolutional network for traffic forecasting

Graph learning-based spatial-temporal graph convolutional neural networks for traffic forecasting

T-GCN: A Temporal Graph Convolutional Network for Traffic Prediction

T-GCN: A Temporal Graph ConvolutionalNetwork for Traffic Prediction

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

Dual Dynamic Spatial-Temporal Graph Convolution Network for Traffic Prediction

Dynamic Adaptive and Adversarial Graph Convolutional Network for Traffic Forecasting

MAGCN: A Multi-Adaptive Graph Convolutional Network for Traffic Forecasting.

Spatial-Temporal Interactive Dynamic Graph Convolution Network for Traffic Forecasting

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

Constructing Geographic and Long-term Temporal Graph for Traffic Forecasting

DDP-GCN: Multi-graph convolutional network for spatiotemporal traffic forecasting

Spatial–Temporal Dynamic Graph Convolutional Network With Interactive Learning for Traffic Forecasting

Hierarchical Graph Convolution Network for Traffic Forecasting

Spatial‐temporal Correlation Graph Convolutional Networks for Traffic Forecasting

Local-Global Spatial-Temporal Graph Convolutional Network for Traffic Flow Forecasting

Graph convolutional networks for traffic forecasting with missing values

Attention Based Spatial-Temporal Graph Convolutional Networks for Traffic Flow Forecasting