meta_mod: Meta Models for Regression (Model Stacking)

In egenn/rtemis: Machine Learning and Visualization

Meta Models for Regression (Model Stacking)

Description

Train a meta model from the output of base learners trained using different learners (algorithms)

Usage

meta_mod(
  x,
  y = NULL,
  x.test = NULL,
  y.test = NULL,
  base.mods = c("mars", "ranger"),
  base.params = vector("list", length(base.mods)),
  base.resample.params = setup.resample(resampler = "kfold", n.resamples = 4),
  meta.mod = "gam",
  meta.params = list(),
  x.name = NULL,
  y.name = NULL,
  save.base.res = TRUE,
  save.base.full = FALSE,
  col = NULL,
  se.lty = 3,
  print.base.plot = FALSE,
  print.plot = TRUE,
  plot.fitted = NULL,
  plot.predicted = NULL,
  plot.theme = rtTheme,
  question = NULL,
  verbose.base.res.mods = FALSE,
  verbose.base.mods = FALSE,
  verbose = TRUE,
  trace = 0,
  base.n.cores = 1,
  n.cores = rtCores,
  save.mod = FALSE,
  outdir = NULL,
  ...
)

Arguments

x	Numeric vector or matrix of features, i.e. independent variables
y	Numeric vector of outcome, i.e. dependent variable
x.test	(Optional) Numeric vector or matrix of validation set features must have set of columns as x
y.test	(Optional) Numeric vector of validation set outcomes
base.mods	Character vector: Two or more base learners. Options: select_learn
base.params	List of length equal to N of base.mods. Each element should be a list of arguments to pass to the corresponding base mod
meta.mod	String. Meta learner. Options: select_learn
x.name	Character: Name for predictor set. (What kind of data is it?)
y.name	Character: Name for outcome
se.lty	How to plot standard errors. If a number, it corresponds to par("lty") line types and is plotted with lines(). If "solid", a transparent polygon is plotted using polygon()
resampler	String. Resampling method to use. Options: "bootstrap", "kfold", "strat.boot", "strat.sub"

Details

This is included mainly for educational purposes.

• Train a set of base learners on resamples of the training set x

• Train a meta learner to map bases' validation set predictions to outcomes

• Train base learners on full training set x

• Use the meta learner to predict test set outcome y.test from testing set (x.test)

Author(s)

E.D. Gennatas

egenn/rtemis documentation built on Dec. 17, 2024, 6:16 p.m.