AttributeError: 'GridSearchCV' object has no attribute 'best_estimator_'

Question

I am using grid search to tune parameters of my models (Random Forest, Linear Regression, etc.). So I save gs objects in grid_searches:

gs = GridSearchCV(model, params, cv=cv, n_jobs=n_jobs,
                  verbose=verbose, scoring="mean_squared_error", refit=refit)
gs.fit(trainX,trainy)
grid_searches[key] = gs

Then I want to access the best estimator for each model in order to make predictions:

def predict(testX, testy, grid_searches):
    keys = models.keys()
    for k in keys:
        print("Predicting with %s." % k)
        yhat = grid_searches[k].best_estimator_.predict(testX)

The error is the following:

AttributeError: 'GridSearchCV' object has no attribute 'best_estimator_'

So how should I make predictions using best models found by Grid Search?

You should just be able to just use grid_searches[k].predict(testX). After you call gs.fit(...), gs should have the optimized parameters (based on the search space). What happens if you try just grid_searches[k].predict(...)? Also, what happens if you add a line under gs.fit(...) with print gs.best_estimator_? — Michael Recachinas
– Michael Recachinas, Commented Dec 9, 2015 at 21:50

Sean Easter · Accepted Answer · 2015-12-10 02:20:47Z

4

It's not clear, from the code excerpt, how you set refit. Per the docs, best_estimator_ is only available when this is True. If False, you should still be able to find the best-performing parameters from grid_scores_, then use them with set_params().

edited Dec 10, 2015 at 2:20

answered Dec 9, 2015 at 23:25

Sean Easter

8898 silver badges17 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

AttributeError: 'GridSearchCV' object has no attribute 'best_estimator_'

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related