摘要
通过分析城市轨道交通客流量的时序特征和RBF神经网络的作用机理,将具有不同时序特征的数据分别用不同的神经网络进行处理,建立了基于客流时序特征的并行加权神经网络模型,并用该模型对北京市城市轨道交通各条线路的客流进行预测.结果表明,各线路客流量预测结果的平均绝对百分误差均在10%以下,小于单个神经网络的预测误差,提高了预测精度.
We propose to use the neural network to process the data with timing characteristics by analyzing urban rail transit passenger flow timing characteristics and the mechanism of RBF neural network.A parallel weighted neural network model based on passenger flow sequence characteristics is developed.We use this model to forecast the Beijing urban rail transit passenger flow and get a better forecast result.Each line＇s traffic prediction error is below 10 ％,which is less than the prediction error of a single neural network so as to improve the prediction accuracy.
出处
《北京交通大学学报》
CAS
CSCD
北大核心
2014年第3期1-6,共6页
JOURNAL OF BEIJING JIAOTONG UNIVERSITY
基金
科技部"科技支撑"计划项目资助(2011BAG01B01-1)
关键词
轨道交通
客流预测
组合预测
神经网络
时序特征
railway traffic
passenger flow forecast
combination forecasting
neural network
timing characteristics
