混合不完备数据的拓展高斯核-支持向量机分类方法

A Expand Gaussian Kernel-Support Vector Machine Classification Method for Incomplete Data Set with Hybrid Value

针对混合不完备数据集,提出一种基于邻域联系度距离拓展高斯核函数的支持向量机分类方法。首先,给出了基于邻域联系度距离拓展高斯核函数的定义;其次,给出基于二次函数逼近的支持向量机SMO训练算法和分类算法;最后,取多个UCI数据集进行了实验分析,通过与填充支持向量机、混合距离支持向量机和风险重构支持向量机分类方法进行比较,结果表明提出的分类方法在不对缺失值作任何处理、不改变支持向量机模型结构与约束条件的情况下,仍然获得了优异的分类效果。

In order to process incomplete data set with hybrid value,a novel classification method based on Expand Gaussian kernel-Support vector machine is proposed.Firstly,based on the connection degree distance function for Expand Gaussian kernel is defined;Secondly,the SMO training algorithm base on quadratic function approximation for support vector machine and classifi⁃cation algorithm are provided.Finally,some experiments are carried out on UCI data sets,and compare to the filling support vector machine,the risk reconstruction support vector machine and the HEOM support vector machine,the experiments show that the ex⁃cellent classification results of proposed classification method which is no need to handle missing values and no need to change SVM model structure and constraints are still obtained.