基于高维复杂数据的变量选择方法研究

Research on Variable Selection Method Based on High-Dimensional Complex Data

摘要: 针对目前大多数基于信息论的线性累加特征选择算法的缺点和不足，将类别变量的信息，引入到对待选特征与已选特征子集的冗余性度量之中，考虑到特征与类别变量之间的对称不确定度，提出了一种新的以信息论为基础的，基于最大相关最小冗余原则的过滤式特征选择方法，并在11个公开的标准数据集上进行了验证，通过与6种其他基于信息论的特征选择方法的结果进行对比，验证了所提出的算法的有效性。

Abstract: In view of the shortcomings and deficiencies of most current linear accumulation feature selection algorithms based on information theory, the information of categorical variables is introduced into the redundancy measure of the feature to be selected and the selected feature subset, taking into account the relationship between the feature and the categorical variable, a new filtering feature selection method based on information theory and the principle of maximum correlation and min-imum redundancy is proposed, and verified on 11 public standard datasets. The results of six other information theory-based feature selection methods are compared to verify the effectiveness of the proposed algorithm.