摘要
针对短时交通流时间序列数据非平稳、非线性的特点,为提高短时交通流的预测精度和收敛速度,提出了一种基于集合经验模态分解(EEMD)和改进型粒子群算法(IPSO)优化的最小二乘支持向量机(LSSVM)的组合预测模型。首先,利用集合经验模态分解方法将短时交通流时间序列样本数据分解为多个本征模函数(IMF)和一个残差项(RES),细化了交通流量的信息,提高了建模的精确度;然后,对分解后的每个分量使用LSSVM并结合IPSO算法进行组合预测,通过选择不同的最优的支持向量机核函数,提高模型精度,通过IPSO算法提高预测效率;最后,将各分量预测值进行叠加作为最终交通流预测值。实验结果表明,EEMD-IPSO-LSSVM组合模型的均方根误差(MSE)比LSSVM模型和PSO-LSSVM模型分别降低了47.4%和24.6%,该组合模型提高了预测精度,并且能够快速地预测交通流时间序列。
In view of the nonlinear and non-stationary characteristics of short-term traffic flow data, this article proposes a combined prediction model based on ensemble empirical mode decomposition(EEMD) and improved particle swarm optimization(IPSO) optimized least squares support vector machine(LSSVM) in order to improve its forecast accuracy and rate of convergence. First of all, the short-term traffic flow time series sample data is decomposed by the EEMD decomposition method to multiple intrinsic mode function(IMF) and one Residual, which refines the traffic flow information and improves the accuracy of modeling;then, each component after decomposition is predicted by using the particle swarm optimization algorithm optimized least squares support vector machine, the model accuracy is improved by selecting different optimal support vector machine kernel functions, and the prediction efficiency is enhanced through improved particle swarm optimization;finally, each component prediction value is superimposed as the final traffic flow prediction value. The experimental results show that the mean square error(MSE) of the EEMD-IPSO-LSSVM combined model compared with the LSSVM model and the PSO-LSSVM model is reduced by 47.4% and 24.6%, respectively. Therefore, the combined model improves prediction accuracy and can quickly predict traffic flow time series.
作者
殷礼胜
唐圣期
李胜
何怡刚
Yin Lisheng;Tang Shengqi;Li Sheng;He Yigang(School of Electrical Engineering and Automation,Hefei University of Technology,Hefei 230009,China)
出处
《电子测量与仪器学报》
CSCD
北大核心
2019年第12期126-133,共8页
Journal of Electronic Measurement and Instrumentation
基金
国家自然科学基金(51577046、61673153)资助项目。
关键词
集合经验模态分解
改进型粒子群算法
最小二乘支持向量机
ensemble empirical mode decomposition
improved particle swarm optimization
least squares support vector machine
作者简介
殷礼胜,1994年于安徽大学获得学士学位,2004年于广西大学获得硕士学位,2007年于重庆大学获得博士学位,现为合肥工业大学副教授,主要研究方向为通讯信道建模,交通流预测。E-mail:yls20000@163.com;通信作者:唐圣期,2017年于青岛理工大学获得学士学位,现为合肥工业大学硕士研究生,主要研究方向为通讯信道建模,交通流预测。E-mail:tsq951024@163.com。
