removed unnecessary calculate_transformation method and implemented instead some checks that the train set is executed first (for a proper transformation depending only on train data)
