Clustering of Interval Valued Data Through Interval Valued Feature Selection: Filter Based Approaches.

In this paper, a problem of selectively choosing a few best interval valued features out of several available features is addressed in an Un-supervised environment. Various models belonging to two categories viz., models which transform interval data to crisp and models which accomplish feature selection through clustering of interval valued features are explored for clustering of interval valued data. Extensive experimentation is conducted on two standard benchmarking datasets using suitable symbolic clustering algorithms. The experimental results show that the approaches presented outperform the state-of-the-art models in terms of correct rand index score and number of features selected.