基于离散化的支持向量机特征选择

Discretization Based Feature Selection for Support Vector Machines

基于Nguyen的粗糙集和布尔推理离散化方法提出一种支持向量机特征选择算法,引入粗糙集的一致度指标控制离散化过程的信息损失,从而删除不相关与冗余的属性,而保留支持向量机所需分类信息。实验结果表明,所提算法提高了SVM分类器的预测精度,缩短了训练时间。

This paper presents a feature selection algorithm for support vector machine based on the rough sets and Boolean reasoning approach put forward by Nguyen. The level of consistency, coined from the rough sets theory, is introduced to measure the information loss during discretization so that irrelative or redundant attributes are eliminated while the necessary information for classification is preserved. Experiment results show that the presented algorithm can improve the prediction accuracy and reduce the training time of support vector machine.