Performance metrics to evaluate classification

Quantify the performance of a classification algorithm. Predictions and truth both have to be binary.

performanceMeasure(pred, obs, perf.method = "f.measure", ...)

Arguments

pred

a logical or numeric, where 0 and FALSE represent control, and, 1 and TRUE represent cases

obs

a logical or numeric, where 0 and FALSE represent control, and, 1 and TRUE represent cases

perf.method

a character, specifying the method to use. Available methods can be accessed using perfMethods

...

additional parameters to methods. see details

Value

a numeric, representing the performance

Details

The F-measure requires the beta parameter which can be specified using f.beta which defaults to 1 thereby computing the F1-measure.

See also

perfMethods

Examples

pred <- sample(0:1, 100, replace = TRUE, prob = c(0.75, 0.25))
obs <- sample(0:1, 100, replace = TRUE, prob = c(0.75, 0.25))

#compute the F1 and F2 scores
f1 <- performanceMeasure(pred, obs)
f2 <- performanceMeasure(pred, obs, f.beta = 2)