Prophet-DCRNN traffic flow prediction method fusing multi-modal information

The invention discloses a Prophet-DCRNN traffic flow prediction method fusing multi-modal information, and belongs to the technical field of traffic flow prediction. Although an existing flow prediction method based on deep learning well captures time-space characteristics of traffic flow, actual urban traffic is affected by factors such as weather, holidays and festivals, and meanwhile, traffic jam tends to occur in severe weather, holidays and festivals. The provides the Prophet-DCRNN traffic flow prediction method fusing multi-modal information to overcome the defects of the prior art. The method uses a Prophet time sequence prediction algorithm to capture holiday effects, uses a DCRNN to capture traffic space-time characteristics, and in addition, based on a stacking-like technology, the Prophet algorithm, the DCRNN algorithm, holiday characteristics and weather information are fused, a hybrid model that finally fuses multi-modal information is obtained, so the accuracy of traffic prediction in festivals