x_data_pos = df_orig_positive.iloc[:,:-3].values
x_data_unl = df_orig_unlabeled.iloc[:,:-3].values
why iloc[:, :-3], the feature cols is iloc[:,:-2]
another question，i think it's better to compare in predict same dataset（that is to say，the original model also predict with same numbers of postive(25%)）
x_data_pos = df_orig_positive.iloc[:,:-3].values
x_data_unl = df_orig_unlabeled.iloc[:,:-3].values
why iloc[:, :-3], the feature cols is iloc[:,:-2]
another question，i think it's better to compare in predict same dataset（that is to say，the original model also predict with same numbers of postive(25%)）