基于粗糙集下近似理论的支持向量机分类方法

Classification Method of SVM Based on Lower Approximations Theory of Rough Set

传统的粗糙集理论不能处理连续属性,而且得到的分类规则大多比较复杂.支持向量机理论能够得到简洁的分类规则,也能处理连续属性,但仅适用与小样本,对大样本数据集有一定的局限性.文章首先提出了针对连续属性的粗糙集下近似理论,使粗糙集理论能够应用到连续属性.基于上述理论以及支持向量机分类方法仅与支持向量有关的特性,提出了一种先由粗糙集进行预处理的支持向量机分类方法.实验表明,该方法在缩短训练时间的基础上,保留了支持向量机方法所需分类信息,提高了分类精度,克服了SVM算法的应用瓶颈.

The traditional rough set theory can not deal with the continuous attribute, and the classifi- cation rules which are obtained from the rough set are mostly complicated. Though SVM can get the concise classification rules and can deal with the continuous attribute, but it can only be used for small sam-pies. This paper presents a SVM classification method based on rough set＇s lower approximations theory and it＇s application in continuous attribute. Experiments results show that the method can preserve the necessary information needed by SVM, improve the prediction accuracy and reduce the training time of support vector machine.