CODEWITHSUNDEEP
scikit-learncross-validationgrid-search
E.Asgari
E.Asgari

Reputation: 315

How to feed predefined k-folds in GridSearchCV

I would like to use GridSearchCV for parameter tuning and evaluation over 10 predefined selected folds of data that I have (as list of lists of data indices).

Does anyone know how to feed GridSearchCV in scikit with 10 lists of predefined test folds?

splits=[ [0,10,9,1,2,..] ,[3,5,7,..],[23,4,34,..]] #len(split)=10

greed_search = GridSearchCV(estimator, param_grid=parameters, cv=splits,scoring=scoring, refit=score, error_score=0, n_jobs=n_jobs)

Upvotes: 1

Views: 479

Answers (1)

Alexey Trofimov
Alexey Trofimov

Reputation: 5027

I think that you need to preprocess your folds a little bit like this:

new_splits = []

for i in range(len(splits)):
    train = [j for i in splits[:i] + splits[i + 1:] for j in i]
    test = splits[i]
    new_splits.append([train, test])

To get not only test parts in iterable, but also train part as well

Upvotes: 2

Related Questions