摘要
提出了一种针对分类属性数据特征选择的新算法。通过给出一种能够直接评价分类属性数据特征选择的评价函数新定义,重新构造能实现分类属性数据信息量、条件互信息、特征之间依赖度定义的计算公式,并在此基础上,提出了一种基于互信息较大相关、较小冗余的特征选择(MRLR)算法。MRLR算法在特征选择时不仅考虑了特征与类标签之间的相关性,而且还考虑了特征之间的冗余性。大量的仿真实验表明,MRLR算法在针对分类属性数据的特征选择时,能获得冗余度小且更具代表性的特征子集,具有较好的高效性和稳定性。
In this paper, a novel feature selection approach based on mutual information called More Relevance Less Redun-dancy（MRLR）algorithm for nominal data is proposed. By reconstructing the computation method of the amount of infor-mation, the conditional mutual information, the dependence between the features so that which can be suitable for compu-tation related the nominal data, and a new definition of the evaluation function of feature selection is given, as well as a new feature selection criterion is used to evaluate the importance of each feature, which takes into account both relevance and redundancy. In MRLR, experimental results show that the relevance and redundancy respectively use mutual informa-tion to measure the dependence of features on the latent class and the dependence between features, and it also enhance the correctness and the effectiveness of MRLR algorithm.
出处
《计算机工程与应用》
CSCD
2014年第16期135-139,共5页
Computer Engineering and Applications
关键词
分类属性数据
特征选择
互信息
nominal data
feature selection
mutual information
