文献类型：期刊文献

中文题名：超限学习机在铁路客运量预测中的应用研究

英文题名：Application research of extreme learning machine in railway passenger volume forecast

作者：刘彩霞[1];方建军[1];刘艳霞[1]

第一作者：刘彩霞

机构：[1]北京联合大学自动化学院

第一机构：北京联合大学城市轨道交通与物流学院

年份：2017

卷号：14

期号：9

起止页码：2013-2019

中文期刊名：铁道科学与工程学报

外文期刊名：Journal of Railway Science and Engineering

收录：CSTPCD;;北大核心:【北大核心2014】；CSCD:【CSCD2017_2018】；

基金：北京市属高等学校高层次人才引进与培养计划资助项目(CIT&TCD20150314);北京市自然科学基金资助项目(4142018)

语种：中文

中文关键词：铁路客运量;神经网络;超限学习机

外文关键词：railway passenger volume; neural network; extreme learning machine

摘要：传统的预测方法对实验对象要求严格、需要人工设置大量参数,导致算法学习速度较慢、预测误差较大等不理想结果。本文引入机器学习中的超限学习机(Extreme Learning Machine,ELM)对铁路客运量进行预测,建立铁路客运量网络预测模型。利用国家统计局公布的1997~2014年铁路客运量数据做了验证,并以2014年的数据为依据对客运量进行月度预测。结果表明:2010~2014年预测值与实际值的平均误差为0.61%,2014年每个月预测值与实际值的误差均低于1.00%。重复的实验证明ELM算法具有很好的泛化能力和鲁棒性,为铁路客运量的预测提供了一种新的工具。
The accurate prediction of railway passenger volume has a very important significance to take effectivemeasures for the railway sector.The traditional prediction methods have some problems that it has strictrequirements for test subjects,and that it need set a number of parameters,leading to slower learning speed andlarge errors.The paper introduces Extreme Learning Machine in machine learning machine to forecast passengervolume,and establishes the network prediction model of railway passenger volume.Some testing work was donebased on the data published by National Bureau of Statistics in China from1997to2014,and forecast thepassenger volume of each month in2014.The results show the average error between predicted value and actualvalue is0.61%from2010to2014,and that the error of each month is less than1.00%in2014.The repeatedexperiment result shows that ELM has good generalization ability and robustness,which provides a new tool forthe prediction of railway passenger volume.