基于类内离散度的最小二乘支持向量机

摘要

支持向量机是一种很流行的机器学习方法，在许多领域都有了广泛的使用。传统的支持向量机模型是寻求类之间的间隔最大化，而忽视了一个重要的信息—样本的类内结构，类内离散度。文中将Fisher判别分析里面的类内离散度引入到最小二乘支持向量机中，提出了基于类内离散度的最小二乘支持向量机模型。并通过核函数将样本映射到高维特征空间，在特征空间中进行样本分类。基于UCI数据库的数据集实验测试表明，基于类内离散度的最小二乘支持向量机提高了分类的准确度。%Support vector machine ( SVM) is a popular machine learning technique,and it has been widely used in many real-world ap-plications. Traditional SVMs aims at seeking the hyperplane that maximizes the margin and ignores an important prior knowledge, the within-class structure. It formulates a Least Square Support Vector Machine ( LSSVM) based on within-class scatter for binary classifi-cation,which incorporates minimum within-class scatter in Fisher Discriminant Analysis ( FDA) into LSSVM. The sample points are mapped to a high dimensional feature space where the samples are classified by using the kernel method. Experiments on four benchmark-ing datasets based on UCI show that the proposed WCS-LSSVM can improve the classification accuracy.