Cluster-Based Rough Set Construction

In many data mining applications, cluster analysis is widely used and its results are expected to be interpretable, comprehensible, and usable. Rough set theory is one of the techniques to induce decision rules and manage inconsistent and incomplete information. This paper proposes a method to construct equivalence classes during the clustering process, isolate outlier points and finally deduce a rough set model from the clustering results. By the rough set model, attribute reduction and decision rule induction can be implemented efficiently and effectively. Experiments on real world data show that our method is useful and robust in handling data with noise.