Synopsis
Sets all predictions to 'unknown' (missing value) if the corresponding confidence is smaller than the specified value.
Description
This operator sets all predictions which do not have a higher confidence than the specified one to "unknown" (missing value). This operator is a quite simple version of the CostBasedThresholdLearner which might be useful in simple binominal classification settings (although it does also work for polynominal classifications).
Input
- example set input: expects: ExampleSetMetaData: #examples: = 0; #attributes: 0
, expects: ExampleSet
Output
- example set output:
- original:
Parameters
- class handling: The mode which defines if all classes are handled equally or if class individual thresholds are set.
- min confidence: The minimal confidence necessary for not setting the prediction to 'unknown'.
- min confidences: A list which defines individual thresholds for classes.