Skip to content
Advertisement

How do I make sure GridSearchCV first does the cross split and then the imputing?

I have a GridSearchCV, with a pipeline that looks something like this:

numeric_transformer = Pipeline(steps=[
    ('imputer', SimpleImputer(strategy='most_frequent')),
    ('scaler', StandardScaler())
])


preprocessor = ColumnTransformer(transformers=[
    ('num', numeric_transformer, numeric_features),
])

clf = Pipeline(steps=[
    ('preprocessor', preprocessor),
    ('classifier', LogisticRegression(solver='lbfgs'))
])  

my GridSearchCV looks like this:

search = GridSearchCV(clf, param_grid, cv = 5, scoring = "roc_auc",error_score=0.0)

with Cross Validation = 5

So, how do I ensure that I split the data first, and then impute in the most frequent?

Advertisement

Answer

GridSearchCV will run roughly like this:

for train_index, val_index in StratifiedKFold(n_splits=5).split(X, y):
    X_train, X_val = X[train_index], X[val_index]
    y_train, y_val = y[train_index], y[val_index]

    clf = Pipeline(steps=[
        ('preprocessor', preprocessor),
        ('classifier', LogisticRegression(solver='lbfgs'))
    ]) 

    clf.fit(X_train, y_train)
    clf.predict(X_val, y_val)

You can be sure that SimpleImputer and StandardScaler will do .fit() and .transform() for each fold.

User contributions licensed under: CC BY-SA
1 People found this is helpful
Advertisement