摘要
为了提高传统方法生成交通拥堵指数(TPI)的准确率,引入一种基于经验模态分解(EMD)与Elman神经网络的组合模型实现交通拥堵指数预测。首先,利用EMD将TPI序列分解为不同时间尺度下的IMF分量和剩余分量;然后,通过偏自相关函数(PACF)计算各分量的滞后期数,以此确定各分量在Elman神经网络中的输入和输出变量;之后,通过上述方法计算出各分量预测值并相加;最后,计算出总预测结果。通过计算结果可知,EMD-PACF-Elman预测方法 3个评价指标(平均绝对误差、均方误差、平均绝对百分误差)的计算结果与单一Elman神经网络模型、EMD-Elman神经网络模型、单一BP神经网络模型、EMD-BP神经网络模型相比都为最低,分别为0.562 4、0.598 9、0.110 7。因此,EMD-PACF-Elman预测方法可以有效地预测TPI,同时也为进一步预测交通拥堵趋势提供了依据。
In order to improve the accuracy of generating traffic performance index(TPI)by traditional methods,a combined model based on(EMD)and Elman neural network is introduced to predict the traffic performance index.First,EMD is applied to decompose the sequence of TPI into IMF component and residual component on different time scales;second,partial autocorrelation function(PACF)is used to calculate the lag period of each component and the input and output variables of each component in Elman neural network are determined;third,the predicted values of each component are obtained and the final prediction result is obtained by sum⁃ming them together.According to the results,the three evaluation indexes(mean square error,mean absolute error and mean absolute percentage error)of EMD-PACF-Elman prediction method are the lowest compared with single Elman neural network model,EMDElman neural network model,single BP neural network model and EMD-BP neural network model,and they are 0.5624,0.5989 and 0.1107 respectively.EMD-PACF-Elman prediction method can effectively predict TPI.It also provides an effective basis for further predicting the trend of traffic congestion.
作者
李振国
徐建新
LI Zhen-guo;XU Jian-xin(Faculty of Science,Kunming University of Science and Technology,Kunming 650500,China;School of Metallurgy and Energy Engineering,Kunming University of Science and Technology,Kunming 650093,China)
出处
《软件导刊》
2020年第11期11-16,共6页
Software Guide
基金
云南省万人计划项目(109720190106)
云南省高层次人才项目(132510978220)。
