Pedestrian volume prediction system capable of simultaneously modeling space-time dependence and daily flow correlation

Abstract

The invention discloses a pedestrian volume prediction system capable of simultaneously modeling space-time dependence and daily flow correlation. The system comprises an ST encoder module, an FR encoder module and a decoder module. According to the invention, the convolutional long-term and short-term memory network is used for simultaneously realizing dependence and time dependence of adjacent regions in space; spatial dependencies in different ranges in a geographic space are captured hierarchically by superposing a convolutional neural network; meanwhile, a complete graph reflecting the flow correlation between all the areas in the whole city range is generated based on the daily flow change mode of the areas, fixed-dimensional vector representation is generated for each area through agraph embedding method, and a pedestrian flow prediction result is generated through a layer of long-short-term memory network and two layers of deconvolution neural networks.

Description

technical field [0001] The present invention relates to a technology in the field of artificial intelligence application, in particular to a people flow forecasting system that simultaneously models time-space dependence and daily flow correlation. Background technique [0002] The human flow forecasting problem is a complex spatio-temporal sequence forecasting problem affected by many factors. Existing research can be roughly divided into three categories. The first type of research uses Recurrent Neural Network (RNN) to capture complex temporal correlations, but in this type of research, the regions are independent of each other, so the correlation between regions is ignored. The second type of research considers the relevance of airspace and uses Convolutional Neural Network (CNN) to solve the problem of human flow prediction. However, CNN cannot capture the change information in the time domain. The third category of research attempts to simultaneously model spatio-temp...

AI Technical Summary

Problems solved by technology

The first type of research uses Recurrent Neural Network (RNN) to capture complex temporal correlations, but in this type of research, the regions are independent of each other, so the correlation between regions is ignored.

The second type of research considers the relevance of airspace and uses Convolutional Neural Network (CNN) to solve the problem of human flow prediction. However, CNN cannot capture the change information in the time domain.

Images