SDPN

Abstract

E-hailing car supply and demand prediction is a long-term but challenging task for E-hailing decision support system and intelligent transportation construction. E-hailing car supply and demand prediction is actually predicting the inflow and outflow of some regions in specific time slot. Accurate e-hailing car supply and demand prediction can improve utilization and scheduling of e-hailing platform and reduce customer waiting time. Existing traffic flow prediction approaches mainly utilize region-based sequence image deep learning models or station-based temporal graph deep learning models to capture spatio-temporal dynamic while we argue that temporal graph deep learning models can be transferred to the region-based case. In this paper, we propose the Supply and Demand Prediction Neural Networks (SDPN), a region-based temporal graph deep learning approach for generalizable scenes. SDPN model integrates structures of Graph Convolutional Neural Network (GCN) and Recurrent Neural Network (RNN) to capture the spatial and temporal dynamics respectively. We evaluate the proposed model on DiDi e-hailing dataset of Haikou City and the experimental studies demonstrate SDPN has achieved superior performance of e-hailing car supply and demand prediction compared with some traditional state-of-art baseline models.