Clustering microarray data using model-based double K-means