基于PersonⅦ核函数的支持向量机及其在化学模式分类中的应用  

Support Vector Machine Based on Person Ⅶ Kernel Function and Its Application in Chemical Pattern Classification

作　　者：郑启富[1] 陈德钊[2] 刘化章[3] 

机构地区：[1]浙江工业大学浙西分校化工系,衢州324000 [2]浙江大学化工系,杭州310027 [3]浙江工业大学化学工程与材料学院,杭州310032

出　　处：《分析化学》2007年第8期1142-1146,共5页Chinese Journal of Analytical Chemistry

基　　金：国家自然科学基金资助项目(No20276063)

摘　　要：化学模式分类问题通常是非线性的,而且比较复杂,难以用经典统计方法建立分类判别模型。以支持向量机(SVM)构建的分类器具有更好的分类性能。对于非线性分类,SVM通过核函数将其映射到高维特征空间中,然后再进行线性分类。因此,核函数往往是决定SVM非线性分类性能的关键。实际应用时,一般通过选择几种核函数,并对其参数进行优化,然后根据分类器的预测性能来决定,训练过程非常耗时,而且结果难以保证最优。为此,采用一种通用性的核函数,即PersonⅦ核函数(PUKF),它可取代目前常用的几种核函数,可避免SVM非线性分类器训练过程的核函数选择问题。本研究将基于PUKF的SVM分类器应用于两个化学模式分类问题,均取得了较好的结果。对于多类分类,设计了一种子分类器的构造方法,它在分类性能保持较好的情况下,简化了多类分类器结构,大大降低了计算量。It is usually complex nonlinear relationship between input pattern vector and output character of chemical object. Comparing with traditional statistical modeling methods, support vector machine （SVM） has advantages at solving chemical pattern classification problems. For solving nonlinear classification problem, the kernel function transforms the nonlinear input space into a high dimensional feature space in which the solution of the problem can be represented as being a linear classification problem. Therefore, in all probabili- ty, the performance of SVM classifier is decided by the kernel function, and choosing a proper kernel function is very important. Whereas the nature of the data is usually unknown, it is very difficult to make, on beforehand, a proper choice out of the possible kernel functions. For this reason, during the model building process, usually more than one kernel is applied to select the one that gives the best prediction performance. Unfortunately, this will lead to a very time-consuming optimization procedure. To circumvent this disadvan- tage, a universal kernel function based on the Pearson Ⅶ function （PUKF） is introduced in this study. PUKF can replace the common kernel functions, and simplifies the training process of SVM nonlinear classifier. SVM classifiers based on PUKF were applied to solve two chemical pattern classification problems, and satisfied results were obtained. In addition, a multi-class classifier was designed. Comparing with the common multi- classifiers, it had a simpler structure , and could reduce the calculation complexity greatly, remaine the classi- fication ability almost.

关 键 词：化学模式分类 支持向量机 PERSON Ⅶ核函数 分类器 

分 类 号：O6-0[理学—化学]