mlr (version 2.19.0)

setThreshold: Set threshold of prediction object.

Description

Set threshold of prediction object for classification or multilabel classification. Creates corresponding discrete class response for the newly set threshold. For binary classification: The positive class is predicted if the probability value exceeds the threshold. For multiclass: Probabilities are divided by corresponding thresholds and the class with maximum resulting value is selected. The result of both are equivalent if in the multi-threshold case the values are greater than 0 and sum to 1. For multilabel classification: A label is predicted (with entry TRUE) if a probability matrix entry exceeds the threshold of the corresponding label.

Usage

setThreshold(pred, threshold)

Value

(Prediction) with changed threshold and corresponding response.

Arguments

pred

(Prediction)
Prediction object.

threshold

(numeric)
Threshold to produce class labels. Has to be a named vector, where names correspond to class labels. Only for binary classification it can be a single numerical threshold for the positive class.

See Also

predict.WrappedModel

Examples

# create task and train learner (LDA)
task = makeClassifTask(data = iris, target = "Species")
lrn = makeLearner("classif.lda", predict.type = "prob")
mod = train(lrn, task)

# predict probabilities and compute performance
pred = predict(mod, newdata = iris)
performance(pred, measures = mmce)
head(as.data.frame(pred))

# adjust threshold and predict probabilities again
threshold = c(setosa = 0.4, versicolor = 0.3, virginica = 0.3)
pred = setThreshold(pred, threshold = threshold)
performance(pred, measures = mmce)
head(as.data.frame(pred))