An improved method of semantic driven subtractive clustering algorithm

摘要

On the basis of SCM (Subtractive Clustering Method), SDSCM is proposed that user semantic concept is quantized by the membership function based on AFS (Axiomatic Fuzzy Sets), and that the quantized user semantic concept is used to automatically determine the density radius T, to semi-automatically determine weight τ. A new index, Semantic Strength Expectation, is brought forward in order to assess the clustering quality. Semantic Strength Expectation along with existed clustering indexes is compared and analyzed among SDSCM, FCM on Wine data set and Iris data set. The analysis results of the experiments show that Semantic Strength Expectation of SDSCM is strongest among three clustering methods.