A Novel Fuzzy Monotone Relationship Method With its Application on Inclusion Degree

Abstract

This paper introduces a new fuzzy monotone relationship and its associated method, which are applied to feature selection and correlation analysis. Specifically, after the concept of a fuzzy monotone is introduced, this paper first defines a new fuzzy monotone relationship between inputs and output. Second, a fuzzy inclusive monotone model is constructed on inclusion degree through several proved propositions, together with presenting a fuzzy inclusive monotone decision membership function. Third, a new algorithm is developed according to the proposed model for feature selection or correlation analysis. Compared with several methods, the proposed algorithm has been validated on several data sets. The results indicate that the proposed algorithm is effective for the selection of numeric attributes, and the correlation analysis. The novel fuzzy monotone relationship and the method are validated through theoretic proof and experimental results.