How can i impelement SMOTE inside a columnTransformer?

Question

I'm trying to implement SMOTENC inside a column transformer. However I'm getting error. The code and the error is provided below.

#Create a mask for categorical features
categorical_feature_mask = X_train.dtypes == object
categorical_columns = X_train.columns[categorical_feature_mask].tolist()
print(categorical_columns)

from imblearn.over_sampling import SMOTENC

#Create two datasets also create a pipeline to automate the preprocessing steps
num_features= X_train.select_dtypes(include=[np.number]).columns
cat_features = X_train.select_dtypes(exclude=[np.number]).columns

cat_transformer = Pipeline(steps=[('imp_c', SimpleImputer(strategy='most_frequent')),
                                  ('label_bina', LabelBinarizer())])
scale_transformer=Pipeline(steps=[('imp_m',SimpleImputer(strategy='median')),
                                  ('std',StandardScaler())])
smote=SMOTENC(categorical_features=categorical_columns,random_state=99)
col_transform = ColumnTransformer(transformers=[
        ('num', scale_transformer, num_features),
        ('cat', cat_transformer, cat_features),
        ('smote', smote )],remainder='passthrough')
#We fit a DecisionTreeClassifier and evaluste the model performance
dt=DecisionTreeClassifier(random_state=99)
pl_dt=Pipeline(steps=[('transform',col_transform),('dt',dt)])
pl_dt.fit(X_train,np.ravel(y_train))

While running this I get error: not enough values to unpack (expected 3, got 2). More precisely

ValueError                                Traceback (most recent call last)
<ipython-input-34-a874d44f98ee> in <module>
      2 dt=DecisionTreeClassifier(random_state=99)
      3 pl_dt=Pipeline(steps=[('transform',col_transform),('dt',dt)])
----> 4 pl_dt.fit(X_train,np.ravel(y_train))
      5 

~/anaconda3/lib/python3.7/site-packages/sklearn/pipeline.py in fit(self, X, y, **fit_params)
    328         """
    329         fit_params_steps = self._check_fit_params(**fit_params)
--> 330         Xt = self._fit(X, y, **fit_params_steps)
    331         with _print_elapsed_time('Pipeline',
    332                                  self._log_message(len(self.steps) - 1)):

~/anaconda3/lib/python3.7/site-packages/sklearn/pipeline.py in _fit(self, X, y, **fit_params_steps)
    294                 message_clsname='Pipeline',
    295                 message=self._log_message(step_idx),
--> 296                 **fit_params_steps[name])
    297             # Replace the transformer of the step with the fitted
    298             # transformer. This is necessary when loading the transformer

~/anaconda3/lib/python3.7/site-packages/joblib/memory.py in __call__(self, *args, **kwargs)
    353 
    354     def __call__(self, *args, **kwargs):
--> 355         return self.func(*args, **kwargs)
    356 
    357     def call_and_shelve(self, *args, **kwargs):

~/anaconda3/lib/python3.7/site-packages/sklearn/pipeline.py in _fit_transform_one(transformer, X, y, weight, message_clsname, message, **fit_params)
    738     with _print_elapsed_time(message_clsname, message):
    739         if hasattr(transformer, 'fit_transform'):
--> 740             res = transformer.fit_transform(X, y, **fit_params)
    741         else:
    742             res = transformer.fit(X, y, **fit_params).transform(X)

~/anaconda3/lib/python3.7/site-packages/sklearn/compose/_column_transformer.py in fit_transform(self, X, y)
    525         # set n_features_in_ attribute
    526         self._check_n_features(X, reset=True)
--> 527         self._validate_transformers()
    528         self._validate_column_callables(X)
    529         self._validate_remainder(X)

~/anaconda3/lib/python3.7/site-packages/sklearn/compose/_column_transformer.py in _validate_transformers(self)
    274             return
    275 
--> 276         names, transformers, _ = zip(*self.transformers)
    277 
    278         # validate names

ValueError: not enough values to unpack (expected 3, got 2)

How can i solve the above error?

Answer 1

ColumnTransformer is used to apply transformations to a subset a columns of the dataset. Since you want to apply SMOTENC to the full dataset, just put it outside the ColumnTransformer. Also, since SMOTENC does not have a fit_transform method, we cannot use it with a scikit-learn pipeline. We need to use a imblearn pipeline:

from imblearn.pipeline import Pipeline
...
smote = SMOTENC(categorical_features=categorical_columns, random_state=99)
col_transform = ColumnTransformer(transformers=[
        ('num', scale_transformer, num_features),
        ('cat', cat_transformer, cat_features)],
        remainder='passthrough')

dt = DecisionTreeClassifier(random_state=99)
pl_dt = Pipeline(steps=[('transform',col_transform), ('smote',smote), ('dt',dt)])
pl_dt.fit(X_train,np.ravel(y_train))