Predicting Citywide Crowd Flows in Irregular Regions Using Multi-View Graph Convolutional Networks

Being able to predict the crowd flows in each and every part of a city, especially irregular regions , is strategically important for traffic control, risk assessment, public safety. However, it very challenging because interactions spatial correlations between different regions. In addition, affected by many factors: i) multiple xmlns:xlink="http://www.w3.org/1999/xlink">temporal correlations among time intervals: closeness, period, trend; ii) complex xmlns:xlink="http://www.w3.org/1999/xlink">external influential weather, events; iii) xmlns:xlink="http://www.w3.org/1999/xlink">meta features: day, day week, so on. this paper, we formulate flow forecasting irregular regions as xmlns:xlink="http://www.w3.org/1999/xlink">spatio-temporal graph (STG) prediction problem which node represents region with time-varying flows. By extending xmlns:xlink="http://www.w3.org/1999/xlink">graph convolution handle information, propose using xmlns:xlink="http://www.w3.org/1999/xlink">spatial graph build xmlns:xlink="http://www.w3.org/1999/xlink">multi-view convolutional network (MVGCN) problem, where views can capture factors mentioned above. We evaluate MVGCN four real-world datasets (taxicabs bikes) extensive experimental results show that our approach outperforms adaptations state-of-the-art methods. And have developed system now be used internally.