基于AIC的RVM核参数选择方法及其应用
    点此下载全文
引用本文:陈小明.基于AIC的RVM核参数选择方法及其应用[J].计算技术与自动化,2016,(2):38-43
摘要点击次数: 695
全文下载次数: 23
作者单位
陈小明 (厦门城市职业学院 电子与信息工程系福建 厦门361008) 
中文摘要:在采用高斯径向基函数的相关向量机(RVM)回归模型中,核参数与模型性能之间关系复杂,针对如何确定RVM核参数的问题,提出一种基于AIC准则选择RVM的核参数的方法。首先基于Akaike Information Criterion (AIC)思想,得出一种新的统计量Q,同时将Q作为适应度函数;然后利用微分进化算法(Differential Evolution Algorithm,DE)对核参数进行寻优,以此选择确定核参数;最后利用该算法建立RVM回归模型对黄金价格进行短期预测。实验结果表明,该模型较传统方法建立的预测模型具有更高的拟合精度和更好的泛化能力,进一步证明基于AIC准则选择RVM的核参数的方法的可行性和有效性。
中文关键词:径向基函数  核参数  相关向量机  微分进化算法  AIC准则  黄金价格
 
Application of RVM Kernel Parameter Selection Algorithm Based on AIC
Abstract:In the Relevance Vector Machine (RVM) regression model based on Gaussian radial basis function, the relationship between kernel parameter and the RVM regression model performance is complex. And in order to determine the RVM kernel parameter, a new fitness function based on the idea of Akaike Information Criterion (AIC) was proposed. Firstly, a new statistical quantity was presented, then, Q acted as fitness function. Secondly, differential evolution algorithm(DE) was used to find the optimal parameters. Finally, gold price in short term was predicted with RVM regression model based on the algorithm. The results show that the model has higher fitting precision and better generalization ability. The effectiveness and availability of RVM kernel parameter selection algorithm based on AIC was also proved.
keywords:Radial Basis Function(RBF)  kernel parameter  Relevance Vector Machine (RVM)  Differential Evolution Algorithm (DE)  Akaike Information Criterion (AIC)  gold price
查看全文   查看/发表评论   下载pdf阅读器