Human Flow Prediction Model Based on Graph Convolutional Recurrent Neural Network

At present, many cities have launched real-time “congestion degree”, “traffic index” and other travel reference indicators, according to which urban residents can reasonably choose to travel. However, these real-time monitoring travel reference indicators do not predict the future, but only respond to the past situation. Therefore, it is necessary to study the prediction of urban crowd flow, and thus become the focus of research. The existing research focuses on dividing cities into grids and predicting the flow of people in each grid area in the next period. However, considering the rapid change of urban crowd flow, measures such as crowd density control require a long response time and the specific location of peak traffic, this paper uses the structure of the graph to divide regions, and innovatively constructs the regional geographic adjacency diagram and regional flow diagram, and uses the graph convolutional cyclic neural network to model the spatial and temporal relationship of regional human flow. In this paper, the public data set of a city taxi is selected for experimental verification and analysis. The experimental results show that the proposed method improves the prediction accuracy of regional passenger flow and reduces the root-mean-square error by 0.04, which is better than other existing schemes.