metric: performance metrics

In linxihui/lazyML: Automatic machine learning algorithms and hyper-parameters selection

Description

built-in performance metrics for classification, regression and survival

Usage

twoClassSummary(data, lev = levels(as.factor(data[, "obs"])), model = NULL,
  metric = availableMetric("twoClass"), ...)

multiClassSummary(data, lev = levels(as.factor(data[, "obs"])),
  model = NULL, metric = availableMetric("multiClass"), ...)

regressionSummary(data, lev = NULL, model,
  metric = availableMetric("regression")[1:4], ...)

survivalSummary(data, lev, model, is.prob = NA, surv.percentile = 0.5,
  max.surv.time = max(data$obs[, 1], na.rm = TRUE),
  metric = availableMetric("survival")[1:3], na.rm = TRUE, ...)

defaultSummary(data, ..., type = c(NA, "twoClass", "multiClass", "regression",
  "survival"))

requireSummary(...)

availableMetric(type = c("twoClass", "multiClass", "regression", "survival"),
  is.prob = TRUE)

Arguments

data	Data frame with columns 'obs', 'pred' and possible columns with class labels as names if class probabilities needed
lev	Class labels for a classification problem. NULL for others
model	Model names (optional)
metric	Performance metrics to be computed
...	Other argument to pass to internal methods
is.prob	TRUE or FALSE indicated if class probability is availabed in data.
surv.percentile	Percentile used to compute quantile survival time as predition if data has time to survival information.
max.surv.time	Maximum follow-up time
na.rm	Should NA rows in data be removed before metric calculation
type	Problem type, either 'twoClass', 'multiClass', 'regression' or 'survival'

Details

availableMetric gives a list of metrics currently supported. defaultSummary is a wrap over twoClassSummary, multiClass, regressionSummary and survivalSummary. It automatically determines which one to use by response in data$obs. requireSummary a variation of modifyFunction to modify defaultSummary, particularly the metric argument to request a vector of non-default metric. See modifyFunction for example

linxihui/lazyML documentation built on May 21, 2019, 6:39 a.m.