摘要
为准确预测地铁车厢的拥挤度,考虑到站时车厢的下车人数、立席面积和车厢承载量等因素,提出一种基于BP神经网络的地铁车厢拥挤度预测方法。基于调查数据和Matlab平台,构建初始BP神经网络、训练以及测试等环节,实现对地铁车厢到站时各车门下车人数的预测。以立席密度为标准进行车厢拥挤度划分,标定即将到站地铁的各节车厢拥挤度。以宁波市鼓楼地铁站为例,对BP神经网络预测方法进行验证,得到不同结构的BP神经网络预测结果。结果表明,最佳预测结果的决定系数R2为0.94,平均相对误差为0.25,预测误差在可控范围内,BP神经网络在下车人数预测上是适用的。
In order to accurately forecast congestion of subway carriages,a method based on BP neural network is proposed considering factors including number of passengers getting off the subway,areas of seats,and capacity of carriages.Based on survey data and Matlab,an initial BP neural network and its training and testing processes are developed to forecast the number of people get off when a subway arrives at a station.Congestion of carriages is divided by density of standing passengers.The congestion of each carriage of a arriving subway can be calibrated.Gulou subway station in Ningbo,China is taken as a case study.The method is verified by obtaining forecast results with different structures of BP neural network.The results show that the optimal forecast has a R 2 of 0.94,and an average relative error of 0.25.The error of forecast is within a controllable range.BP neural network is applicable to forecast the number of people get off the subway.
作者
方晨晨
周继彪
董升
王依婷
陈莎雯
FANG Chenchen;ZHOU Jibiao;DONG Sheng;WANG Yiting;CHEN Shawen(School of Transportation,Wuhan University of Technology,Wuhan 430063,China;School of Civil and Transportation Engineering,Ningbo University of Technology,Ningbo 315211,Zhejiang,China;School of Transportation Engineering,Tongji University,Shanghai 201804,China)
出处
《交通信息与安全》
CSCD
北大核心
2018年第6期47-53,共7页
Journal of Transport Information and Safety
基金
浙江省哲学社会科学规划课题(18NDJC107YB)
浙江省大学生科技创新项目(2017R424023)
宁波市自然科学基金(2018A610127)资助
关键词
交通工程
车厢拥挤度
BP神经网络
立席密度
预测方法
traffic engineering
congestion of subway carriages
BP neural network
density of standing passengers
forecast method
