Model-based methods to identify multiple cluster structures in a data set