The invention provides a dynamic graph
convolution traffic speed prediction method which comprises the steps: step 1, matching vehicle
GPS trajectory data into an
urban road network, and obtaining a
traffic speed time sequence of each road section; step 2, regarding road sections of the
urban road network as graph nodes, regarding intersections of the
urban road network as connecting edges of a graph, constructing a road network graph, and obtaining an adjacent matrix between the road sections; step 3, calculating the
traffic speed similarity between adjacent road sections according to the traffic speed
time sequence of each road section, and obtaining a real-time adjacent road section
similarity matrix; and step 4, inputting the traffic speed
time sequence of each road section and the adjacent road section
similarity matrix into a graph
convolution network for training to obtain a future road section traffic speed prediction result. According to the invention, spatial dependence and time dependence between road sections can be learned in real time, the change rule of the traffic speed can be captured, the speed of future urban roads can be predicted more accurately, and the methodcan be applied to intelligent traffic and
smart city construction.