基于MCC的自适应混沌序列预测算法仿真  

MCC Based Adaptive Prediction Algorithm Simulation for Chaotic Series

作　　者：马文涛[1] 张蒙[1] 赵芳玲[1] 

机构地区：[1]西安航空学院基础部,陕西西安710077

出　　处：《计算机仿真》2013年第3期247-250,355,共5页Computer Simulation

摘　　要：针对基于LMS的自适应预测算法对具有时变特性的时间序列预测在鲁棒性等方面存在缺陷,而使用最大广义相关熵准则以衡量输入输出的相似度,它含有误差分布的高阶统计量,对数据处理具有一定的鲁棒性,提出了一种基于MCC的混沌时间序列自适应预测算法,考虑到LMS算法和MCC准则的优势,将输入序列和权值向量分成两组,分别用LMS和MCC进行迭代训练,得到组合的新自适应预测算法。仿真结果表明,组合自适应预测算法在预测精度和鲁棒性方面都要优于基于LMS或基于MCC的预测算法。The adaptive prediction algorithm, based on the LMS for the time series prediction of time-varying characteristics, has the defect in the aspect of robustness. The maximum Correntropy criterion is used to measure the similarity between input and output which contain the high order statistics of the error distribution. The paper pro- posed a chaotic time series prediction algorithm based on the MCC. And considering advantages of LMS algorithm and MCC, the input sequence and weight vectors were divided into two groups, iterative training based on the LMS and MCC respectively, to obtain the new combination of adaptive prediction algorithm. The simulation experiment shows that the combination adaptive prediction algorithm in prediction accuracy and robustness is superior to LMS and MCC based prediction algorithms.

关 键 词：最大广义相关熵准则 最小均方 组合 混沌时间序列预测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]