A distance metric suitable for fuzzy partitioning

摘要

In this paper, we address the problem to build up hum data a fuzzy partition in an iterative procedure based on an original distance metric. Thanks to the used distances: internal distance within a fuzzy set, external distance between two fuzzy sets, distance between prototypes, the family of generated fuzzy partitions keeps the relevant properties of legibility and interpretability. A small and academic example, the iris data, demonstrates the efficiency of the proposed approach.