Mechanisms of partial supervision in rough clustering approaches

Abstract

We bring two rough-set-based clustering algorithms into the framework of partially supervised clustering. A mechanism of partial supervision relying on either qualitative or quantitative information about memberships of patterns to clusters is envisioned. Allowing such knowledge-based hints to play an active role in the clustering process has proved to be highly beneficial, according to our empirical results. Other existing rough clustering techniques can successfully incorporate this type of auxiliary information with little computational effort.