We reduced the training set to 80% and then to 60% and we observed what will happen to the classification rate.