近似马尔科夫毯混合式特征选择

Hybrid feature selection algorithm based on approximate Markov blanket

基于Filter-Wrapper模型混合式特征选择,先利用近似马尔科夫毯删除冗余和不相关的特征,再用SBS算法对过滤后的特征集进行选择,筛选出最佳特征子集。在2个不同的分类器和4个公开数据集上与两种特征选择算法做了对比试验。

Based on Filter-Wrapper model hybrid feature extraction,the redundant and irrelevant features are deleted by means of approximate Markov blanket.Then SBS algorithm is used to select the filtered feature set and screen out the best feature subset.The method is compared with other two methods on two classifiers and 4 public data sets.