基于结构风险最小化的加权偏最小二乘法

为了在偏最小二乘法(PLS)建模过程中实现结构风险最小化(SRM)，提出基于结构风险最小化的加权偏最小二乘法(WPLS)。WPLS先提取训练样本中的主元，然后使用支持向量机(SVM)训练算法计算训练样本权值，最后计算原始论域中的回归模型。该算法保留了PLS能有效地提取对系统解释性最强的信息的优点，并通过样本权值提高模型的泛化能力，从而实现SRM准则，所建立的模型具有可解释性。仿真计算证明了模型的有效性。

Structure risk minimization based weighted partial least-squared method

Weighted Partial Least-Squared (WPLS) method was proposed to achieve Structure Risk Minimization(SRM)in the Partial Least-Squares (PLS) modeling process.At first,WPLS abstracted the principal components of training samples,and then it trained the weight of samples by means of Support Vector Machine (SVM) algorithm,and finally computed the regression model in the original universe discourse.WPLS not only takes the advantage of PLS to extract most explanatory variables,but also improves generalization property through the weight of samples and SRM is achieved with interpretable model.Simulation results show the effectiveness of the proposed method.