Improving fuzzy C-means clustering algorithm based on a density-induced distance measure

Abstract

The authors report an improved fuzzy C-means algorithm in comparison with the conventional one by employing a density-induced distance metric based on a novel calculation method of relative density degree. By using various synthetic and real data sets, the clustering performance of the proposed method is systematically studied and compared with that of the conventional one. The obtained results support the conclusion that this novel method does not only inherit good characteristics of the traditional one, but also possesses improved partitions.

Keywords

• fuzzy set theory
• pattern clustering
• synthetic data sets
• real data sets
• relative density degree
• density-induced distance measure
• fuzzy C-means clustering algorithm