Deep Learning for Spatio-Temporal Sequence Forecasting: A Survey

In this paper, the latest progress of deep learning models in the application of spatio-temporal sequences prediction was summarized. First, the attributes and types of spatio-temporal sequence data, as well as the corresponding instantiation and representation, were introduced. Then, to deal with the three problems existed in spatio-temporal sequences, different data preprocessing methods were proposed, and the prediction methods based on traditional parameter models, traditional machine learning models and deep learning models were illustrated and compared, respectively, which provided a guidance for researchers to select proper models. Moreover, the application of deep learning models to spatio-temporal sequence prediction in different fields was depicted. Finally, the current research deficiencies and the future research directions for the spatio-temporal sequence prediction were suggested.