Soft clustering from crisp clustering using granulation for mobile call mining

This paper builds on an earlier study for transforming clustering schemes between two levels of granularity. A crisp clustering scheme of coarse granules can be easily transferred to a crisp clustering at finer level of granularity. However, crisp clustering of finer granules cannot be easily transferred to coarser granules due to the non-overlapping nature of crisp clusters. We propose the use of soft clustering schemes, which allow for possible overlap of clusters. The clustering of finer granules given by phone calls is first transferred to fuzzy clustering and then to rough clustering. The process highlights the complementary nature of fuzzy and rough set theory.