Research And Application Of Spatial-temporal Dynamic Graph Attention Network Based Sharing Travel Vehicle Usage Demand Prediction

As a novel travel mode,sharing travel plays a more and more important role in the travel service market and intelligent transportation system,which provides people more convenience in their daily travel.Sharing travel vehicle usage demand prediction is an effective solution to the mismatch problem between the distribution of vehicles and user demand in sharing travel systems.An accurate demand prediction model can help platforms pre-allocate resources to improve vehicle utilization and user experience.As an essential task in spatial-temporal data mining,sharing travel vehicle usage demand prediction has attracted extensive attention of researchers in the field of intelligent transportation.Recent related works mainly focus on Graph Convolutional Networks(GCN)to model the complicated irregular non-Euclidean spatial correlations.However,existing GCN-based sharing travel vehicle usage demand prediction methods only assign the same importance to different neighbor regions,and maintain a fixed graph structure with static spatial relationships throughout the timeline when extracting the irregular non-Euclidean spatial correlations.This paper proposes the Spatial-Temporal Dynamic Graph Attention Network(STDGAT),a novel sharing travel vehicle usage demand prediction method.Based on the attention mechanism of GAT,STDGAT extracts different pair-wise correlations to achieve the adaptive importance allocation for different neighbor regions.STDGAT constructs a novel time-specific commuting-based dynamic spatial graph structure to capture the dynamic time-specific spatial relationships throughout the timeline.Extensive experiments are conducted on a real-world sharing travel dataset,and the experimental results demonstrate the significant improvement of our method on three evaluation metrics RMSE,MAPE and MAE over state-of-the-art baselines.Finally,this paper designs and implements a sharing travel vehicle usage demand prediction prototype system based on the proposed STDGAT.The main work of this paper can be summarized as follows:(1)This paper introduces the background and significance of sharing travel vehicle usage demand prediction,and summarizes recent related works.(2)This paper proposes the STDGAT,a novel sharing travel vehicle usage demand prediction method.Based on the GAT,STDGAT achieves adaptive importance assignment for different neighbor regions.(3)This paper proposes a time-specific commuting-based dynamic spatial graph structure.By this way,STDGAT can capture the dynamic time-specific spatial features in different time intervals throughout the timeline.(4)This paper conducts extensive experiments on a large-scale real-world sharing travel dataset.The experimental results show the superiorities of our method on three evaluation metrics RMSE,MAPE and MAE,compared with state-of-the-art baselines.(5)Based on the proposed STDGAT,a sharing travel vehicle usage demand prediction system prototype is designed and implemented.