一种基于粗糙集理论的支持向量机分类算法

粗糙集理论和支持向量机在数据挖掘方面具有较强的互补特性,基于粗糙集理论的上近似集、下近似集和边界域概念,结合支持向量机的分类原理,提出了一种支持向量机分类算法。首先,在支持向量机分类中定义样本分类的粗糙集规则,然后在边界域寻找两类样本中使判别式绝对值取值最小且分类正确的样本来确定最优分类面,脱离了对惩罚系数C的寻优问题,有效避免了过拟合问题,并通过循环迭代算法寻找合适的参数b,获得分类性能更优的支持向量机,最后通过对一个二维样本数据库进行分类实验,验证了此算法的有效性与可行性。

A Support Vector Machine Classification Algorithm Based on Rough Sets Theory

Rough sets theory(RST) and support vector machine(SVM) has strong complementary characteristics in terms of data mining.A classification algorithm based on lower approximation set,upper approximation set and the boundary region in RST and classification principle in SVM has been presented.Rough sets rules of sample classification based on SVM classification are defined.Then the sample in the margin region which makes the absolute value of the discriminant minimization is searched and the correct classification to obtain the optimal separating surface has been gotten.It gets out of the optimization problem of the penalty coefficient C,avoids the over-fitting problem,and searches the most optimal coefficient b by cyclic iterative method to obtain a better SVM.Finally,a classification experiment on a two-dimensional sample database proves the validity and feasibility of this algorithm. Support Vector Machine(SVM);Rough Sets;Classification;Penalty Coefficient