A Full Traffic Forecasting Method Based on Dual Graph Framework

Abstract

The invention discloses a full-scale traffic prediction method based on a dual graph framework, including: (1) expressing the road network structure as a topological graph, using intersections as nodes, and road sections connecting the intersections as edges; preparing historical edge and node data and future edge and node data; (2) Construct a historical information encoder, input historical data into the encoder, realize information transmission between edges and nodes through multi-layer dual mapping, and output multi-layer dual mapping Splicing into historical feature tensors; (3) constructing a future prediction decoder, decoding historical feature tensors into future spatio-temporal features, and outputting future prediction results; (4) using the error between the prediction results and actual data as a loss function Model training until the loss function converges; (5) Use the trained model to test the model, and apply it after the test is completed. The prediction result of the present invention can obtain a full and complete description of the future traffic conditions, and the prediction accuracy is high.

Description

technical field [0001] The invention belongs to the field of artificial intelligence and time series forecasting, and in particular relates to a full traffic forecasting method based on a dual graph framework. Background technique [0002] The traffic forecasting task refers to predicting the traffic data of a future period based on the traffic data of a period of history within the range of a given regional road network, including traffic flow, average speed, and travel time prediction. Traffic prediction is the core of Intelligent Transportation System (ITS), and it is widely used in traffic efficiency optimization, traffic risk control, route planning and navigation, etc. Traffic forecasting is a very challenging problem. The difficulty lies in how to model the complex spatiotemporal dependencies between different locations in the road network. [0003] Traffic forecasting can be classified as a time series forecasting problem with road network structure, which has been ...

AI Technical Summary

Problems solved by technology

[0006] The present invention provides a full-volume traffic prediction method based on the dual graph framework, which well solves the problem of missing edge prediction results in existing traffic prediction methods; even if only node prediction is made, because the edge historical data, the present invention can also obtain better prediction accuracy

Images