Cumulative Gains Chart — gain • modelblueprint

Plots cumulative gains curves for one or more competing scores against a perfect model baseline. The Gini coefficient for each score is shown in the legend.

Usage

gain(data, ...)

# S3 method for class 'modelblueprint'
gain(
  data,
  set = c("train", "test", "holdout"),
  title = NULL,
  ret = c("plot", "data", "gini"),
  ...,
  precomputed_preds = NULL
)

Arguments

data: A modelblueprint object.
...: Passed to the default method.
set: [character] Dataset splits to use: any of "train", "test", "holdout". Defaults to all available (non-NULL) sets. When more than one set is used, a named list with one result per set is returned.
title: Chart title. Defaults to model_display_name (with the set name appended when plotting multiple sets).
ret: "plot", "data", or "gini". Default "plot".
precomputed_preds: [numeric | NULL] Optional vector of pre-computed predictions (one per row of the requested set). When supplied, the internal predict.modelblueprint() call is skipped. Use this in loops or dashboards where predictions have already been computed to avoid redundant scoring.

Value

A plotly object, list of data.tables, or list of Gini values.

Examples

# \donttest{
mb <- modelblueprint(
  model  = glm(vs ~ wt + hp, data = mtcars, family = binomial),
  train  = mtcars,
  y_name = "vs",
  model_display_name = "logistic_vs"
)
gain(mb)

# }