Learn R Programming
rtemis (version 0.79)

varSelect: Variable Selection by Variable Importace

Description

Select important variables from a set of features based on RANGER- or XGBLIN-estimated variable importance

Usage

varSelect(x, y, method = c("RANGER", "XGBLIN"), xgb.params = list(alpha
  = 0.1, lambda = 0.1), p = 0.2, print.plot = TRUE, verbose = TRUE)

Arguments

x

Matrix / Data Frame of Predictors

y

Outcome vector

method

String: "RANGER", "XGBLIN": Learner to use for estimating variable importace. Default = "RANGER"

xgb.params

List of parameters for method = "XGBLIN"

p

Float (0, 1): Fraction of variables in x to select. p * ncol(x). May help to set to a fraction twice what you expect to be the true fraction of useful variables, to reduce false negatives at the expense of false positives which can be dealt by an appropriate learning algorithm. (Default = .2)

print.plot

Logical: If TRUE, print index plot of variable importance using mplot3.x

verbose

Logical: If TRUE, print messages to screen