Hierarchical Graph Convolution Network for Traffic Forecasting

Traffic forecasting is attracting considerable interest due to its widespread application in intelligent transportation systems. Given the complex and dynamic traffic data, many methods focus on how establish a spatial-temporal model express non-stationary patterns. Recently, latest Graph Convolution Network (GCN) has been introduced learn spatial features while time neural networks are used temporal features. These GCN based obtain state-of-the-art performance. However, current ignore natural hierarchical structure of systems which composed micro layers road macro region networks, nodes obtained through pooling method could include some hot regions such as downtown CBD etc., only applied graph networks. In this paper, we propose novel Hierarchical Networks (HGCN) for by operating both graphs. The proposed evaluated two city speed datasets. Compared like WaveNet, HGCN gets higher precision with lower computational cost.The website code https://github.com/guokan987/HGCN.git.