Dynamic Spatio-temporal Graph-based CNNs for Traffic Prediction

• DOI: 10.48550/arxiv.1812.02019
• Publication Date: 2018-01-01
• Authors: Ken Chen; Fei Chen; Baisheng Lai; Zhongming Jin; Yong Liu; Kai Li; Long Wei; Pengfei Wang; Yandong Tang; Jianqiang Huang; Xian‐Sheng Hua
• Keywords: FOS: Computer and information sciences; Computer Vision and Pattern Recognition (cs.CV); Computer Science - Computer Vision and Pattern Recognition; 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Forecasting future traffic flows from previous ones is a challenging problem because of their complex and dynamic nature spatio-temporal structures. Most existing graph-based CNNs attempt to capture the static relations while largely neglecting dynamics underlying sequential data. In this paper, we present (DST-GCNNs) by learning expressive features represent structures predict surveillance video particular, DST-GCNN two stream network. flow prediction stream, novel convolutional layer extract graph representation flows. Then several such layers are stacked together over time. Meanwhile, between in often time variant as condition changes To dynamics, use structures, predicted fed into stream. Experiments on real datasets demonstrate that proposed model achieves competitive performances compared with other state-of-the-art methods.