The invention provides a
traffic station flow prediction method based on space-time multi-graph
convolution. The
traffic station flow prediction method is used for solving the problems that in the prior art, the feature capture capacity and prediction precision of
traffic station flow prediction are not high. The traffic
station flow prediction method comprises the following steps: firstly, constructing a neighbor graph and a circulation flow graph, respectively constructing
convolution components and capturing spatial and temporal feature output of
station flow , mapping the spatial and temporal feature output into flow values with the same shape as a to-be-predicted result, and fusing the two components to obtain a spatial and temporal multi-graph
convolution network model based on context gating; constructing training and testing data according to the
station access flow data, obtaining a mature space-time multi-graph convolutional
network model, and completing station flow prediction is completed. According to the method, multi-graph convolution is applied to
deep mining of traffic station flow data, spatial-temporal characteristics of traffic station flow are fully captured from spatial dimensions and time dimensions, various factors used for predicting traffic station in-out flow are comprehensively considered, and traffic station flow prediction precision is improved.