毕业设计（论文）-非线性分类问题的研究汇.docVIP

摘 要 本文论述了支持向量机的基本原理和思想，侧重分析了支持向量机的两种多类分类训练算法（OAA算法、OAO算法）和四种核函数（即线性内核、径向基函数内核、多项式内核，S型内核）。本论文实现了三种算法（还包括BSVM算法），并使用这三种训练算法和这四种核函数实现了非线性模式分析模型。 本论文利用matlab程序语言，实现了一个图形界面。该界面能使用户灵活的选择三种训练算法的一种和四个核函数的一个对数据集进行分类，并能手动创建数据和装载已有数据，还能清除已创建或装载的数据。对于分类结果，图形界面还能显示出分类所用的训练算法，核函数，算法训练时间，支持向量个数以及分类错误率。根据算法训练时间，支持向量个数和分类错误率的统计，本文还做成三个表格，简单分析了三种算法的优劣。 本论文的创新点在于：原有算法只能使用一种核函数和一种C调整系数来修正算法的分类效果，本论文通过收集核函数的资料，扩展到能使用四种核函数，并修改少量代码使算法能使用两种C调整系数。 关键词：支持向量机，多类分类，核函数 Abstract This text has expounded the basic principle and thought of Support Vector Machine, and has laid particular emphasi on analyzing two kinds of Multi-class classification training algorithms (OAA algorithm, OAO algorithm) and four kernel functions(Linear, Polynomial, RBF, sigmoid) of Support Vector Machine. This thesis implement the three training algorithms（include , BSVM algorithm）, and then, implement the non_linear pattern analysis model. This thesis implement a graphical interfaces using matlab program language. It is convenient to choose one of the three training algorithms and one of the three kernel functions to classify the data by the graphical interfaces. And the graphical interfaces implement the function of creating data sets by user and loading the existent data sets. Moreover, the graphical interfaces can clear the data sets which is created and loaded by user. For the result of classification, the graphical interfaces can display the sort of training algorithms and kernel functions, the time algorithm training, the number of support vector and the error ratio of classification. This text have made three table by the stat of the time algorithm training, the number of support vector and the error ratio of classification, and analyse simply the quality of the three training algorithms. This thesis have some innovation. The former algorithm can use a sort of kernel function and one kind of C parameter. And now, the algorithm can use four kernel functions by collecting information about kernel function and