Finding Interesting Subspace Clusters From High Dimensional Datasets