Flow prediction method and device

The embodiment of the invention provides a traffic prediction method and device. The method can comprise the steps that features of a point location to be predicted in multiple dimensions are acquired, and the features in the multiple dimensions at least comprise topological graph structure features and historical traffic features; the features of the to-be-predicted point location in the multiple dimensions are input into a flow prediction model, a flow prediction result of the to-be-predicted point location is obtained, the flow prediction model is trained by sample data marked with true values in advance, and the flow prediction result of the to-be-predicted point location is obtained. The sample data is the features, marked with true values, of multiple point locations in the to-be-predicted point location and the other point locations in the multiple dimensions. Therefore, the traffic prediction model does not discretely learn the mapping from the features of different point locations to the traffic any