Irregular region flow prediction method based on multi-graph convolution and GRU

Abstract

The invention discloses an irregular region flow prediction method based on multi-graph convolution and GRU. The method comprises the following steps: 1, dividing a region into N disconnected irregular regions; 2, performing space-time simplification on historical trajectory data, and performing calculating to obtain the inflow and outflow of all regions under each time step; 3, establishing a plurality of association graphs among the regions, constructing a corresponding adjacent matrix, and representing diversified spatial associations among the irregular regions; 4, designing a multi-graphconvolution neural network based on the association graph between the regions, and fusing the diversified spatial association features between the regions to obtain a multi-graph convolution fusion result; 5, capturing time correlation by adopting a GRU neural network based on a multi-graph convolution fusion result; 6, selecting a proper loss function, performing training to obtain a prediction model, and predicting through the prediction model to obtain the inflow amount and the outflow amount of each region.

Description

technical field [0001] The invention relates to the field of traffic flow forecasting, in particular to a method for forecasting traffic in irregular areas based on multi-image convolution and GRU. Background technique [0002] Traffic flow prediction is an important part of intelligent transportation system. The purpose of regional flow forecasting is to predict the future flow value in urban areas based on given historical data. Accurate forecasting can help traffic managers to control and manage flow in advance. [0003] Regional flow forecasting methods usually exploit spatial and temporal correlations between regions. Traditional regional flow forecasting uses time series forecasting methods, such as autoregressive moving average model (ARIMA), time-varying Poisson model, and vector autoregressive model. They only consider time-dependent associations and have low prediction accuracy. With the rise of deep learning, researchers use deep learning models to predict traf...

AI Technical Summary

Problems solved by technology

However, these methods can only predict traffic in regular regions

Images