摘要
针对属性特别多仅用一种属性约简方法难以实现有效约简的情况,提出了基于双重属性约简的混合支持向量机分类方法.通过引入贡献率和正确率两个概念,首先采用主成分分析算法计算各个条件属性的贡献率,根据贡献率大小和给定的阈值去掉条件属性中贡献率小的成分,提取信息量最大的主要成分;然后再基于粗糙集的属性约简理论,计算这些主要成分对决策变量的正确率,对这些属性进行第二次约简;该方法采用定性定量相结合的方式,可以最大程度地去除属性集中冗余的或不重要的属性,保证将最简的属性样本集输入支持向量机进行建模预测.最后的仿真试验验证了我们所提方法的有效性和正确性.
A hybrid support vector machine classification algorithm based on dual attribute reductions is proposed on the condition that it is difficult to achieve effective reduction to numerous attributes only by one algorithm of attribute reduction. In this algorithm, two definitions about the selection of attribute importance - correlation degree and contribution degree are introduced. Firstly, correlation degree of each condition attribute is computed based on principal component analysis （PCA） and then the most informative principal components are selected through deleting secondary components based on the correlation degree and the given threshold. Secondly, the contribution degree of selected attributes in the process of PCA to decision variances is computed and then the attribute reduction is implemented again on the basis of rough sets. In the hybrid algorithm, the combination of qualitative and quantitative is applied in order to delete redundant and unimportant attributes and input the fewest attribute sample sets into SVM to modeling prediction. Finally, the simulation experiments show the effectiveness and accuracy of the suggested hybrid method.
出处
《昆明理工大学学报（理工版）》
2006年第4期113-117,共5页
Journal of Kunming University of Science and Technology(Natural Science Edition)
关键词
粗糙集
支持向量机
混合分类算法
rough sets
support vector machines
dual attribute reduction
