Urban traffic prediction method based on hypergraph neural network under structural data missing

The invention discloses an urban traffic prediction method based on a hypergraph neural network under structural data loss. The urban traffic prediction method is composed of four main parts: data input, hypergraph construction, feature propagation, hypergraph convolution and traffic prediction. Based on multi-mode static topographic data and dynamic traffic data, two semantic hypergraphs and one geospatial graph are constructed and integrated to describe high-order semantic association and second-order geospatial association of fine granularity and coarse granularity of traffic conditions. In order to learn feature representation on a hypergraph, a new hypergraph convolution operator is obtained from graph convolution and hypergraph learning theories. The proposed hypergraph convolution is used as a deep network of a construction module, advanced feature representation of prediction is learned, and then the traffic condition is predicted. According to the method, feature propagation is carried out based on t