A Novel Membership Function Definition for Fuzzy Classification

Abstract

In this paper, a novel membership function is defined for fuzzy sets by using supervised learning approach. Firstly, in a supervised learning approach, training dataset is separated with the previously defined polyhedral conic functions. Then obtained polyhedral conic functions are used for defining a new membership function. After that by using this function a new fuzzy classification algorithm is defined to classify fuzzy sets that have the similar structure. The algorithm with all suggested methods is implemented on real-world datasets and the performance values are compared with the state of art classification algorithms.