A Double-Clustering Approach for Interpretable Granulation of Data