Feature Selection Based on Correlation between Fuzzy Features and Optimal Fuzzy-Valued Feature Subset Selection

Feature selection plays an important role in classification or recognition. The aim of feature selection is to reduce the number of features used in classification. In the whole feature space, there might be strong correlation between the features. Feature selection based on information theory is proposed for avoiding redundant features. However, such algorithm only focuses on the case that the feature values are discrete. This paper proposes a method includes correlation between features based on fuzzifying the numeric-value features. In this paper, we suggest a method of constructing compact feature space before feature selection. It aims at removing redundant features which may be correlative with some other features in the original feature space and improvements in classification performance.