AUTHOR=Li Cong , Zhang Huyin , Wang Zengkai , Wu Yonghao , Yang Fei 

TITLE=Spatial-Temporal Attention Mechanism and Graph Convolutional Networks for Destination Prediction

JOURNAL=Frontiers in Neurorobotics

VOLUME=Volume 16 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/fnbot.2022.925210

DOI=10.3389/fnbot.2022.925210

ISSN=1662-5218

ABSTRACT=Urban transportation destination prediction is a crucial issue in the area of intelligent transportation, which plays an important role in urban traffic planning, traffic congestion management, and intelligent traffic control and management. Because destination prediction is limited to highly nonlinear and complex spatial road networks, it is not only spatio-temporal dependence but also has a certain loop periodicity, which is one of the challenging topics in recent years. Existing destination prediction methods has limited ability to model large-scale spatial data that changes dynamically with time, so they cannot obtain satisfactory prediction results. This paper proposes a humanin-loop spatial-temporal attention mechanism with graph convolutional network (STAGCN) model combined to solve the problem of destination prediction. STAGCN is mainly composed of three main parts, namely GCN module, LSTM module and attention module, which respectively model the space and time attributes. The model divides large-scale urban traffic networks into grids, and GCN module is used to capture dynamic spatial-temporal correlations in traffic data. Using a spatio-temporal attention mechanism for the analysis of features with loop periodicity and enhancing the features of key nodes in the grid, and then the spatial and temporal features are combined as the input of LSTM to produce the final prediction results. The proposed model is evaluated on a large scale urban real data set. Experimental results show that compared with some traditional baseline models, STAGCN model has achieved better performance in urban car-hailing destination prediction.