A Simplified Deep Residual Network for Citywide Crowd Flows Prediction

Abstract

Crowd flows prediction is an important problem of urban computing. The existing best-known method adopts deep residual networks to model spatio-temporal properties and often achieves good prediction performance. However, since three separated network structures are used to model the properties, the time cost is often expensive for the best-known method. In this paper, we propose an improved method to reduce the running time of the best-known method by simplifying its architecture. Compared with the best-known method, the training time and predicting time of our method can be reduced dramatically. Moreover, the improved method can achieve similar prediction performance with the best-known method. Extensive experiments on the real-world datasets were conducted to show the efficiency of our proposed method.