[Python] How to use GridSearchCV with a custom scoring metric in a classification problem?

I want to use GridSearchCV to tune hyperparameters for a classifier, but I also want to use a custom scoring function that gives higher weight to recall. How can I implement this properly in scikit-learn?

from sklearn.metrics import make_scorer, recall_score

def custom_score(y_true, y_pred):
return recall_score(y_true, y_pred) * 1.5

scorer = make_scorer(custom_score)


Is this the correct way to pass a weighted scoring function?
Do I need to define greater_is_better?

Continue reading...