To choose an appropriate number of class, , we have used two clustering validity measures – the Silhouette coefficient [15] and the Davies-Bouldin index [16]. We applied the kmeans classification on the training data by varying the number of classes and calculated the two measures for each . Table III, shows the best number of classes for the two measures. It’s shows that value, 6 is selected as the best number of class by all two clustering evaluation measures and second best was 5 and 3 by both separately.