A Multi-View Spatial-Temporal Network for Vehicle Refueling Demand Inference

Petroleum station location selection is an important component to enabling public service in a smart city. While refueling data from petroleum stations is collected, vehicle refueling demand inference involves many other external features, such as POIs, road networks, and meteorological data. Traditional location selection approaches mostly rely on discrete features, which fail to model the complex non-linear spatial-temporal relations. We leverage both the information from petroleum stations and urban data that are closely related to refueling demand inference and design an end-to-end structure based on unique properties of spatio-temporal data. More specifically, we employ feedforward FC layers and LSTMs for modeling spatial and temporal features as well as capturing deep feature interactions. Experiments on real-world vehicle refueling dataset demonstrate the effectiveness of our approach over the state-of-the-art methods.