Multi-perspective convolutional neural networks for citywide crowd flow prediction

Crowd flow prediction is an important problem of urban computing with many applications, such as public security. Inspired by the success of deep learning, various deep learning models have been proposed to solve this problem. Although existing methods have achieved good prediction performance, they cannot effectively capture richer spatial-temporal correlations that are important for crowd flow prediction. To address the limitation of existing methods, we propose a novel 2D CNN-based (convolutional neural networks) model via multiple perspectives called the MPCNN to capture richer spatial-temporal correlations. In particular, three perspective CNNs are included in the MPCNN: the front CNN, the side CNN and the top CNN. Then, we propose a fusion layer to combine the results of the three CNNs. In addition, in the MPCNN, we use external factors to enhance prediction performance. Based on four real-world datasets, we performed a series of experiments to compare the proposed method with existing methods, and experimental results demonstrate the effectiveness and efficiency of the proposed method.