A spatio-temporal long-term and short-term urban pedestrian flow prediction method based on deep learning

Abstract

The invention discloses a spatio-temporal long-term and short-term urban human flow prediction method based on deep learning, which is used for the human flow prediction. When extracting the spatial correlation, the present invention performs local convolution on the correlation of adjacent regions, and extracts the correlation of remote regions by graph convolution, which reduces the parameters and calculation amount while taking into account the integrity of the spatial correlation; the present invention simultaneously Captures short-term time dependence and long-term time dependence, making prediction results more accurate in the time dimension; the present invention takes into account the distribution of regional semantic information, and assigns corresponding proportions to each type of POI in each region Weights, which more precisely exploit the influence of regional semantics on urban pedestrian flow.

Description

technical field [0001] The present invention relates to the field of human flow forecasting, in particular to a spatio-temporal long-term and short-term urban human flow forecasting method based on deep learning. Background technique [0002] With the advancement of urbanization, developed urban roads and diverse modes of transportation have brought great convenience to people's travel. However, if the traffic data cannot be used reasonably, the effect will often be counterproductive, followed by problems such as traffic congestion, waste of resources, and potential safety hazards. [0003] For example, road congestion in big cities, frequent traffic accidents, and the New Year's Eve stampede on the Bund in Shanghai have sounded the alarm for urban traffic flow control. The meaning of urban traffic flow prediction is to predict the traffic flow at the current moment or the next moment based on historical traffic flow data. Its significance is that, in view of the predicted...

AI Technical Summary

Problems solved by technology

The existing technology considers the dynamic correlation of inter-regional human flow and the periodic drift in time series, and solves the long-term time dependence problem, but ignores the long-distance spatial dependence, and only considers one or several influencing factors. To predict people flow information well, it is necessary to comprehensively consider various influencing factors, including spatio-temporal features, especially the long-distance spatial dependence and long-term time dependence, regional semantic features and additional features.

The existing technology is still blank for the research of comprehensive consideration of various influencing factors

Images