Urban OD (Origin-Destination) people flow prediction method based on gravity multi-layer three-dimensional residual network

Abstract

The invention discloses an urban OD (Origin-Destination) people flow prediction method based on a gravity multi-layer three-dimensional residual network. Firstly, three OD information inputs are constructed, and then a multilayer three-dimensional residual network based on gravitational mechanics is trained; the multi-layer three-dimensional residual network based on gravitational mechanics comprises a background information module, a plane OD historical information module and a gravitational mechanics-based OD historical information module. Shallow layer features and deep layer features of plane and gravitational mechanics-based OD historical information are subjected to jumping type aggregation connection; a cyclic three-dimensional tensor vertical and horizontal self-attention block is designed and is used for learning the influence of time-space-background information of local and global neighbors on crowd flow; and finally, urban OD people flow prediction can be carried out by using the trained network. The OD passenger flow in the future time period obtained through prediction can be used for pre-judging passenger flows of different traffic modes, such as online car-hailing, taxis, public transportation, shared bicycles and other traffic systems, and platform benefits and service quality are balanced.

Description

technical field [0001] The invention belongs to the field of urban crowd flow prediction, and in particular relates to an urban OD crowd flow prediction method based on a multi-layer three-dimensional residual network of gravity. Background technique [0002] As urban populations grow, different modes of transportation such as taxis, shared cars, and public transportation play an important role in many cities. Due to the ever-increasing passenger demand and the acquisition of real-time data of people flow by various sensors, people are more and more interested in using data to implement dynamic and adaptive strategies before the surge of travel demand in order to prevent the degradation of service quality. [0003] Predicting people flow demand can be used to (1) better balance the spatial distribution of vehicle re-resources; (2) reduce vehicle idle time, improve system service quality, and achieve higher platform profits. But most currently focus on regional traffic-level...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the challenge of city-wide OD crowd flow prediction, the present invention provides an urban OD crowd flow prediction method based on gravity multi-layer 3D residual network

This method can accurately and effectively learn the space-time dependence relationship between urban global regions with time evolution, and solve the noise effect caused by data sparseness.

Images