Spatial–temporal attention fusion for traffic speed prediction

Accurate vehicle speed prediction is of great significance to the urban traffic intelligent control system. However, in terms prediction, modules that integrate temporal and spatial features existing methods are effective short-term but medium-term or long-term errors relatively large. In order reduce predict speed, this paper proposes a method combines attention Spatial–temporal features, referred as ASTCN. Specifically, unlike previous methods, ASTCN can use convolutional network (ATCN) separately extract from collected by each sensor, mechanism then perform spatial–temporal feature fusion. Experiments on three real-world datasets show proposed model outperforms state-of-the-art baselines.