Setting Threshold for Predict

[korzh]

Suppose we have some binary classification task and train the model using any of binary classifier available (LogisticRegressionBinaryClassifier or FastTreeBinaryClassifier or anything else).

Then we evaluate our model on a testing set using BinaryClassificationEvaluator. During this evaluation we can set a threshold using Threshold property of the evaluator.

The question is: how can we set the threshold for prediction?

[Ivanidzo4ka]

@Zruty0 I can't find way to do that in new API as well, is it supported?

[Zruty0]

This is the closest:

machinelearning/test/Microsoft.ML.Tests/Scenarios/Api/Estimators/ReconfigurablePrediction.cs

Line 41 in a285f8d

var newModel = new BinaryPredictionTransformer<IPredictorProducing<float>>(ml, model.Model, trainData.Schema, model.FeatureColumn, threshold: 0.01f, thresholdColumn: DefaultColumnNames.Probability);

[Ivanidzo4ka]

Marking as up-for-grabs and good first issue, in case someone want to update cook book examples.