Traffic flow prediction method based on combination of multiple attention and adaptive graph convolution

The invention discloses a traffic flow prediction method based on multi-head attention combined with adaptive graph convolution, and is applied to the technical field of traffic flow prediction. Comprising the following steps: S1, constructing a model; s2, modeling the spatial correlation of the traffic flow by using graph convolution, capturing adjacent spatial information of each node on a graph, and extracting spatial features of the traffic flow; s3, discovering the connection of hidden nodes in the graph structure from the data through training, thereby capturing the spatial relationship of the hidden traffic flow in the data; s4, capturing time dependence information of adjacent time slices of the traffic flow; s5, extracting spatio-temporal information of different spatial dimensions from different historical moments by introducing a plurality of attention heads, and capturing dynamic spatio-temporal correlation of traffic flow to obtain high-dimensional spatio-temporal features; and S6, fusion is carr