Research On A Feature Selection Based On Intuitionistic Fuzzy Entropy

With the advent of the era of big data,the "dimensional disaster" has become a problem that plagues researchers.Faced with: "dimensional disaster",in order to be able to analyze and mine data efficiently and accurately,researchers currently use feature selection methods to reduce the data.With the deepening of research,many feature selection algorithms have appeared,but various methods have their limitations and defects to a certain extent.In order to efficiently and accurately obtain the optimal feature subset and achieve higher performance in data analysis and mining,this paper proposes a feature selection algorithm based on intuitionistic fuzzy entropy.First,the feature selection algorithm proposed in the paper uses the intuitionistic fuzzy C-means clustering algorithm to obtain the intuition membership of each feature data to the class target.Secondly,the intuitionistic fuzzy entropy corresponding to the feature is calculated by the obtained intuition membership,and finally the entropy The size of the value is used as an evaluation measure for feature selection.The smaller the intuitionistic fuzzy entropy of the feature,the greater the recognition degree of the feature,the greater the contribution to data classification,and then the feature selection of the original data is completed.Feature selection of the 20-newsgroups high-dimensional data set through experiments to obtain the optimal feature subset,and then use the optimal feature subset to construct KNN and SVM classifiers,and then use the classifier to classify the 20-newsgroups data set to classify The results are evaluated by using the classifier to evaluate the accuracy of the index,and the F1 score of the harmonic average of the accuracy and recall.It is verified by experiments that the proposed algorithm performs better on feature selection of high-dimensional feature data and is superior to general feature selection algorithms.In general,the feature selection algorithm based on intuitionistic fuzzy entropy researched in this paper solves the problem of inefficient classifiers in processing highdimensional data sets.The contributions of the paper are as follows: 1.Through the combination of intuitionistic fuzzy set and information entropy theory,The membership function and the non-membership function are used to characterize the influence of the intuitionistic fuzzy set uncertainty on the intuitionistic fuzzy entropy.Therefore,the description of the ambiguity of the feature on the target is used as the evaluation measure of feature selection;2.The construction of the intuitionistic membership function At present,there is no unified mathematical model.This paper proposes a method for generating membership functions using intuitionistic fuzzy C-means clustering algorithm.