A performance measure is evaluated after a single train/predict step and returns a single number to assess the quality of the prediction (or maybe only the model, think AIC). The measure itself knows whether it wants to be minimized or maximized and for what tasks it is applicable.

All supported measures can be found by listMeasures or as a table in the tutorial appendix: https://mlr.mlr-org.com/articles/tutorial/measures.html.

If you want a measure for a misclassification cost matrix, look at makeCostMeasure. If you want to implement your own measure, look at makeMeasure.

Most measures can directly be accessed via the function named after the scheme measureX (e.g. measureSSE).

For clustering measures, we compact the predicted cluster IDs such that they form a continuous series starting with 1. If this is not the case, some of the measures will generate warnings.

Some measure have parameters. Their defaults are set in the constructor makeMeasure and can be overwritten using setMeasurePars.

featperc

timetrain

timepredict

timeboth

sse

measureSSE(truth, response)

mse

measureMSE(truth, response)

rmse

measureRMSE(truth, response)

medse

measureMEDSE(truth, response)

sae

measureSAE(truth, response)

mae

measureMAE(truth, response)

medae

measureMEDAE(truth, response)

rsq

measureRSQ(truth, response)

expvar

measureEXPVAR(truth, response)

arsq

rrse

measureRRSE(truth, response)

rae

measureRAE(truth, response)

mape

measureMAPE(truth, response)

msle

measureMSLE(truth, response)

rmsle

measureRMSLE(truth, response)

kendalltau

measureKendallTau(truth, response)

spearmanrho

measureSpearmanRho(truth, response)

mmce

measureMMCE(truth, response)

acc

measureACC(truth, response)

ber

measureBER(truth, response)

multiclass.aunu

measureAUNU(probabilities, truth)

multiclass.aunp

measureAUNP(probabilities, truth)

multiclass.au1u

measureAU1U(probabilities, truth)

multiclass.au1p

measureAU1P(probabilities, truth)

multiclass.brier

measureMulticlassBrier(probabilities, truth)

logloss

measureLogloss(probabilities, truth)

ssr

measureSSR(probabilities, truth)

qsr

measureQSR(probabilities, truth)

lsr

measureLSR(probabilities, truth)

kappa

measureKAPPA(truth, response)

wkappa

measureWKAPPA(truth, response)

auc

measureAUC(probabilities, truth, negative, positive)

brier

measureBrier(probabilities, truth, negative, positive)

brier.scaled

measureBrierScaled(probabilities, truth, negative, positive)

bac

measureBAC(truth, response)

tp

measureTP(truth, response, positive)

tn

measureTN(truth, response, negative)

fp

measureFP(truth, response, positive)

fn

measureFN(truth, response, negative)

tpr

measureTPR(truth, response, positive)

tnr

measureTNR(truth, response, negative)

fpr

measureFPR(truth, response, negative, positive)

fnr

measureFNR(truth, response, negative, positive)

ppv

measurePPV(truth, response, positive, probabilities = NULL)

npv

measureNPV(truth, response, negative)

fdr

measureFDR(truth, response, positive)

mcc

measureMCC(truth, response, negative, positive)

f1

measureF1(truth, response, positive)

gmean

measureGMEAN(truth, response, negative, positive)

gpr

measureGPR(truth, response, positive)

multilabel.hamloss

measureMultilabelHamloss(truth, response)

multilabel.subset01

measureMultilabelSubset01(truth, response)

multilabel.f1

measureMultilabelF1(truth, response)

multilabel.acc

measureMultilabelACC(truth, response)

multilabel.ppv

measureMultilabelPPV(truth, response)

multilabel.tpr

measureMultilabelTPR(truth, response)

cindex

cindex.uno

iauc.uno

ibrier

meancosts

mcp

db

dunn

G1

G2

silhouette

Arguments

truth

(factor)
Vector of the true class.

response

(factor)
Vector of the predicted class.

probabilities

(numeric | matrix)
a) For purely binary classification measures: The predicted probabilities for the positive class as a numeric vector. b) For multiclass classification measures: The predicted probabilities for all classes, always as a numeric matrix, where columns are named with class labels.

negative

(character(1))
The name of the negative class.

positive

(character(1))
The name of the positive class.

Format

none

References

He, H. & Garcia, E. A. (2009) Learning from Imbalanced Data. IEEE Transactions on Knowledge and Data Engineering, vol. 21, no. 9. pp. 1263-1284.

H. Uno et al. On the C-statistics for Evaluating Overall Adequacy of Risk Prediction Procedures with Censored Survival Data Statistics in medicine. 2011;30(10):1105-1117. https://doi.org/10.1002/sim.4154.

H. Uno et al. Evaluating Prediction Rules for T-Year Survivors with Censored Regression Models Journal of the American Statistical Association 102, no. 478 (2007): 527-37. https://www.jstor.org/stable/27639883.

See also