基于循环神经网络的城市轨道交通短时客流预测  

Short-term passenger flow forecasting of urban rail transit based on recurrent neural network

作　　者：张惠臻[1] 高正凯 李建强 王晨曦 潘玉彪 王成[1] 王靖[1] ZHANG Hui-zhen;GAO Zheng-kai;LI Jian-qiang;WANG Chen-xi;PAN Yu-biao;WANG Cheng;WANG Jing(School of Computer Science and Technology,Huaqiao University,Xiamen 361021,China;School of Software Engineering,Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China;Linewell Software Co.,Ltd.,Quanzhou 362000,China)

机构地区：[1]华侨大学计算机科学与技术学院,福建厦门361021 [2]北京工业大学信息学部软件学院,北京100124 [3]南威软件股份有限公司,福建泉州362000

出　　处：《吉林大学学报（工学版）》2023年第2期430-438,共9页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金项目(61802133);福建省科技计划重点项目(2020H0016,2021H0019)。

摘　　要：为更好地预测城市轨道交通的短时客流情况,提出了基于循环神经网络模型的预测方法。首先,针对轨道交通进出站客流数据,利用Pearson相关系数确定短时客流影响因素;然后,改进K-means聚类算法划分高、中、低客流量三类轨道站点,分析客流时空分布规律及高峰时间段;最后,采用分别基于长短时记忆神经网络(LSTM)与门控循环单元(GRU)的短时客流预测方法,预测不同类型站点在不同时段的客流。实验结果表明:5 min为预测的最佳时间粒度,在此时间粒度下GRU模型整体性能优于LSTM模型。In order to better predict the short-term passenger flow of urban rail transit, a prediction method based on the recurrent neural network model is proposed. Firstly, based on the actual passenger flow data of each rail transit station, the Pearson correlation coefficient is used to determine the influencing factors of short-term passenger flow of rail transit, such as the weather conditions, historical passenger flow, whether it is a peak time period, whether it is a working day, etc. Secondly, the K-means clustering algorithm is used to classify rail transit stations into three types: high, medium, and low passenger flow stations. Then the distribution of passenger flow for each station type in time and space is analyzed, to determine the peak period of passenger flow for each station type. Finally, two urban rail transit short-term passenger flow prediction methods based on long-short term memory neural network(LSTM) and gated recurrent unit(GRU) respectively are proposed to predict the passenger flow of each type of station in different time period. The experimental results show that 5 min is the best time granularity for short-term passenger flow prediction of the two models. In this time granularity, the overall performance of the GRU model is better than the LSTM model.

关 键 词：交通运输工程 城市轨道交通 短时客流预测 循环神经网络 长短时记忆神经网络 门控循环单元 

分 类 号：U121[交通运输工程]