Short-Term Passenger Flow Prediction for Urban Rail Transit Based on Time-Space Attention Graph Convolutional Network

Abstract

How to predict the future subway passenger flow based on the flow of multiple stations is a huge challenge. It has been proved by researchers that graph neural networks have more advantages than methods that only make predictions for a single node. In this paper, we propose a space-time attention network, which is an improved graph convolutional neural network based on the attention mechanism. This space-time attention network extends the attention mechanism to the time dimension of the node. Our method surpasses the prediction effect of the latest graph convolutional network, and it also achieves better results on Hangzhou subway passenger flow data than the suboptimal model.