Hi , I have been using your GaussianNB package for machine learning modeling by naive bayes algorithm. I have generated train and test set using scikit learn (set1). Then I modified that train and test set (set2) in a controlled way for my experiment. The classification should perform similarly (should return same class label for set1 and set2) according to the theory but it is behaving otherwise. Can you give me any insights why is this happening? Thanks, prashanta saha, PhD student, Montana State University,Bozeman,MT.