Spatiotemporal Virtual Graph Convolution Network for Key Origin-Destination Flow Prediction in Metro System

Short-term Origin-Destination (OD) flow prediction plays a major part in the realization of Smart Metro. It can help traffic managers implement dynamic control strategies to improve operation safety. Also, it assist passengers making reasonable travel plans passenger experience. However, there are problems that dimension OD short-term is much higher than base number metro stations and matrix sparse. To resolve above two problems, threshold-based method proposed extract key pairs first. contains attribute information station exhibits similar time evolution characteristics, so spatial temporal correlation needs be considered prediction. Pearson used build virtual graph model connection between pairs. A spatiotemporal convolutional network (ST-VGCN), which combines advantages neural gated recurrent network, identify associations patterns simultaneously. The evaluated on 39 days real-world data from Shenzhen Metro, outperforms other benchmarks. research this work contribute development forecasts provide ideas for real-time management rail transit. Furthermore, establish early warning mechanisms quickly evacuate large case emergency.