Multi-Task-Based Spatio-Temporal Generative Inference Network for Predicting Highway Traffic Speed.

Vehicles and traffic management departments have a strong desire to know the traffic speed on the high-way network in a certain future time. A multi-task-based spatio-temporal generative inference network (MT-STGIN) is proposed to predict highway traffic speed in this paper. MT-STGIN can handle the following three challenges: (1) dynamic spatio-temporal correlations, (2) prediction error propagation elimination, and (3) traffic speed heterogeneity on the highway network. First, the encoder is used to extract the dynamic spatio-temporal correlations of the highway network. Second, the decoder concentrates on correlating historical and target sequences and generates the target hidden outputs rather than a dynamic step-by-step decoding way. Finally, a multi-task learning method is used to predict traffic speed on different types of roads because of heterogeneity and shares the underlying network parameters. The evaluation experiments demonstrate that the performance of the proposed prediction model is better than that of the baselines, which are conducted based on the monitoring data of the highway in Yinchuan City, Ningxia Province, China.