Measures (metrics) for statistics and machine learning
Overview
This package defines common measures (metrics) for classification and regression problems in statistics and machine learning. To see if your favorite measure is implemented, see this list. Some multi-target measures are included, but see also Custom multi-target measures.
Measures with parameters (e.g., the $L^p$ loss) are realized as callable instances of a struct; calling syntax complies with the specification in StatisticalMeasuresBase.jl.
In addition to the measures themselves, this package provides:
A tool
roc_curve
for plotting Receiver Operator CharacteristicsAn extension module allowing measures from LossFunctions.jl to be used and extended using the same syntax as other measures. See Using losses from LossFunctions.jl.
A submodule
ConfusionMatrices
providing a confusion matrix type and basic functionality.A submodule
Functions
where some core measure implementations are factored out as pure functions.