Setting the scoring method for hyperopt-sklearn

Question

Is there a way to define the scoring method when using hyperopt-sklearn for classifiers ?

By default, it seems to be the model.score method of sklearn, so accuracy for classifiers. But for some problems (eg : imbalanced classes), accuracy is not the best metric…

score 1 · Answer 1 · answered Nov 13 '19 at 13:36

1

HyperoptEstimator takes an optional loss function argument that defaults to accuracy, see here:

https://github.com/hyperopt/hyperopt-sklearn/blob/master/hpsklearn/estimator.py#L477-L482

answered Nov 13 '19 at 13:36

Max Pumperla

81
2

Would it be possible to provide a working exemple using a custom loss function ? – Krystof Nov 13 '19 at 16:30

Setting the scoring method for hyperopt-sklearn

1 Answers1