R/calculateROCMeasures.R
Calculate the absolute number of correct/incorrect classifications and the following evaluation measures:
tpr
True positive rate (Sensitivity, Recall)
fpr
False positive rate (Fallout)
fnr
False negative rate (Miss rate)
tnr
True negative rate (Specificity)
ppv
Positive predictive value (Precision)
for
False omission rate
lrp
Positive likelihood ratio (LR+)
fdr
False discovery rate
npv
Negative predictive value
acc
Accuracy
lrm
Negative likelihood ratio (LR)
dor
Diagnostic odds ratio
For details on the used measures see measures and also https://en.wikipedia.org/wiki/Receiver_operating_characteristic.
The element for the false omission rate in the resulting object is not called for
but
fomr
since for
should never be used as a variable name in an object.
calculateROCMeasures(pred) # S3 method for ROCMeasures print(x, abbreviations = TRUE, digits = 2, ...)
pred  (Prediction) 

x  ( 
abbreviations  ( 
digits  ( 
... 

(ROCMeasures
).
A list containing two elements confusion.matrix
which is
the 2 times 2 confusion matrix of absolute frequencies and measures
, a list of the above mentioned measures.
print
:
Other roc: asROCRPrediction
Other performance: ConfusionMatrix
,
calculateConfusionMatrix
,
estimateRelativeOverfitting
,
makeCostMeasure
,
makeCustomResampledMeasure
,
makeMeasure
, measures
,
performance
, setAggregation
,
setMeasurePars
lrn = makeLearner("classif.rpart", predict.type = "prob") fit = train(lrn, sonar.task) pred = predict(fit, task = sonar.task) calculateROCMeasures(pred)#> predicted #> true M R #> M 95 16 tpr: 0.86 fnr: 0.14 #> R 10 87 fpr: 0.1 tnr: 0.9 #> ppv: 0.9 for: 0.16 lrp: 8.3 acc: 0.88 #> fdr: 0.1 npv: 0.84 lrm: 0.16 dor: 51.66 #> #> #> Abbreviations: #> tpr  True positive rate (Sensitivity, Recall) #> fpr  False positive rate (Fallout) #> fnr  False negative rate (Miss rate) #> tnr  True negative rate (Specificity) #> ppv  Positive predictive value (Precision) #> for  False omission rate #> lrp  Positive likelihood ratio (LR+) #> fdr  False discovery rate #> npv  Negative predictive value #> acc  Accuracy #> lrm  Negative likelihood ratio (LR) #> dor  Diagnostic odds ratio