Three-way Decision Feature Selection And Classification Methods Under Covering Rough Set

Three-way decision theory has been widely used in many disciplines because it is closer to human cognitive and decision-making.Three-way decision is based on decision-theoretic rough set and probability rough set,which has some limitation.Decision-theoretic rough set and probability rough set are partitioned by equivalent classes of Pawlak rough set.The equivalent classes are hard partitions of data and ignoring the diversity of samples,which lead to the reduction of the decision accuracy of the three-way decision.Aiming at this problem,this paper proposed the model of three-way decision under covering rough set.Covering rough set extends the equivalent classes of Pawlak rough set to covering classes,it allows intersection between each equivalent class,takes full account of the diversity of the samples,and the model of three-way decision under covering rough set is used to deal with feature selection and classification tasks.For feature selection tasks,this paper proposed three-way decision feature selection algorithm under covering rough set.Based on the original covering,this algorithm proposed a new covering,which made the original non-monotonous reduction of three-way decision become monotonous,it's easy to select feature subset with positive region.Firstly,calculated the whole covering under all conditional attributes and calculated the covering operator of each sample under different features.Secondly,calculated the conditional entropy of each sample covering class under different features,and selected the feature with the largest entropy value.Then calculated the significance of every feature and selected the feature with the largest significance value.Finally,the stopping condition was based on whether the number of original positive region was equal to the number of positive region of the selected feature subset.Compared with the traditional feature selection algorithm,the experimental results of the proposed algorithm are better.For classification tasks,this paper proposed three-way decision binary classification algorithm under covering rough set.This algorithm improved the original three-way decision classification algorithm,and classified the boundary region samples using covering operators in covering rough set.This algorithm was divided into three steps.The first step was to use the probability that the sample belongs to each decision class to classify.If the first step could not judge the label of test sample,went to second step.The second step was to use the cover operator of the sample to classify.If the second step could not judge label,went to third step.The third step was to use the distance-based nearest neighbor classification method to classify the sample.Compared with the traditional classification algorithm,the experimental results of the proposed algorithm are better.In this paper,contrast experiments of feature selection and classification algorithms were carried out under the same conditions.Different data sets and different evaluation indicators were used to evaluate these algorithms.Experimental results show that the proposed method are superior to other similar algorithms in most cases.