Optimizes the threshold of predictions based on probabilities. Works for classification and multilabel tasks. Uses BBmisc::optimizeSubInts for normal binary class problems and GenSA::GenSA for multiclass and multilabel problems.
Usage
tuneThreshold(pred, measure, task, model, nsub = 20L, control = list())Arguments
- pred
(Prediction)
Prediction object.- measure
(Measure)
Performance measure to optimize. Default is the default measure for the task.- task
(Task)
Learning task. Rarely neeeded, only when required for the performance measure.- model
(WrappedModel)
Fitted model. Rarely neeeded, only when required for the performance measure.- nsub
(
integer(1))
Passed to BBmisc::optimizeSubInts for 2class problems. Default is 20.- control
(list)
Control object for GenSA::GenSA when used. Default is empty list.
Value
(list). A named list with with the following components:
th is the optimal threshold, perf the performance value.
See also
Other tune:
TuneControl,
getNestedTuneResultsOptPathDf(),
getNestedTuneResultsX(),
getResamplingIndices(),
getTuneResult(),
makeModelMultiplexerParamSet(),
makeModelMultiplexer(),
makeTuneControlCMAES(),
makeTuneControlDesign(),
makeTuneControlGenSA(),
makeTuneControlGrid(),
makeTuneControlIrace(),
makeTuneControlMBO(),
makeTuneControlRandom(),
makeTuneWrapper(),
tuneParams()