基于图卷积网络的乘客打车需求预测  

Passenger Ride-hailing Demand Prediction Based on Graph Convolutional Networks

作　　者：董成祥[1] 魏昕[1] 张坤鹏 汪永超 杨宇辉 DONG Chengxiang;WEI Xin;ZHANG Kunpeng;WANG Yongchao;YANG Yuhui(School of Electromechanical Engineering,Guangdong University of Technology,Guangzhou 510006,China;College of Electrical Engineering,Henan University of Technology,Zhengzhou 450001,China;School of Intelligent Manufacturing,Guangzhou Panyu Polytechnic,Guangzhou 511483,China)

机构地区：[1]广东工业大学机电工程学院,广东广州510006 [2]河南工业大学电气工程学院,河南郑州450001 [3]广州番禺职业技术学院智能制造学院,广东广州511483

出　　处：《工业工程》2022年第5期98-105,共8页Industrial Engineering Journal

基　　金：国家自然科学基金资助项目(62002101)。

摘　　要：乘客打车需求具有非线性和动态性的特点。为了提高乘客需求预测的准确性,需要充分考虑城市区域间的时空特性。针对城市中乘客的打车需求预测问题,利用图卷积网络(graph convolutional network,GCN)和长短期记忆单元(long short-term memory,LSTM)建立GCN-LSTM预测模型。在分析城市区域间时空特性的基础上,基于动态时间规整算法(dynamic time warping,DTW),将具有相似时空特性的区域重组并构建乘客打车需求图,利用图卷积网络提取需求图的空间特征;运用基于LSTM的编码器捕捉区域的时间特性;运用基于LSTM的解码器实现乘客打车需求的多区域同时、多时间步长预测。通过与传统模型的对比实验表明,本文提出的GCN-LSTM预测模型的均方根误差、平均绝对百分比误差和平均绝对误差最小,验证了所提出模型的预测准确性。Predicting passenger demand is challenging due to nonlinear and dynamic demand patterns.The spatiotemporal dependencies among various zones need to be taken into consideration to improve the prediction accuracy.To address the problem of passenger demand prediction,a GCN-LSTM model is established by combining the graph convolutional network(GCN)and long short-term memory(LSTM).Based on the analysis of the spatiotemporal correlation among various urban areas,passenger demand maps are constructed with the help of the dynamic time warping(DTW)algorithm and the spatial dependencies of the demand map captured using GCN.The LSTM-based encoder is used to capture the temporal dependencies of the zones.The LSTM-based decoder is used to carry out the multi-step predictions of passenger demand for multiple zones simultaneously.The comparison experiments show that the proposed GCN-LSTM model outperforms the traditional models in terms of the root mean square error(RMSE),mean absolute percentage error(MAPE),and mean absolute error(MAE),which verifies the prediction accuracy of the proposed model.

关 键 词：图卷积网络 长短期记忆单元 时空特性 乘客打车需求预测 

分 类 号：U491[交通运输工程—交通运输规划与管理]