multi_metric: Multiple Metrics

In MetricsWeighted: Weighted Metrics and Performance Measures for Machine Learning

Multiple Metrics

Description

Provides a way to create a list of metrics/performance measures from a parametrized function like the Tweedie deviance or the elementary scoring functions for expectiles.

Usage

multi_metric(fun, ...)

Arguments

fun	A metric/performance measure with additional parameter to be varied.
...	Further arguments passed to fun(), including one varying parameter (specified by a vector).

Value

A named list of functions.

See Also

performance()

Examples

data <- data.frame(act = 1:10, pred = c(1:9, 12))
multi <- multi_metric(fun = deviance_tweedie, tweedie_p = c(0, seq(1, 3, by = 0.1)))
performance(data, actual = "act", predicted = "pred", metrics = multi)
multi <- multi_metric(
  fun = r_squared,
  deviance_function = deviance_tweedie, tweedie_p = c(0, seq(1, 3, by = 0.1))
)
performance(data, actual = "act", predicted = "pred", metrics = multi)
multi <- multi_metric(fun = elementary_score_expectile, theta = 1:11, alpha = 0.1)
performance(data, actual = "act", predicted = "pred", metrics = multi, key = "theta")
multi <- multi_metric(fun = elementary_score_expectile, theta = 1:11, alpha = 0.5)
performance(data, actual = "act", predicted = "pred", metrics = multi, key = "theta")

MetricsWeighted documentation built on Nov. 16, 2023, 5:09 p.m.