feature.boruta: Boruta Feature Selection - Wrapper

In MarcoNiemann/kaggle_house: Analysing House Prices in Ames, Iowa and Build a Sales Price Prediction Model

Description

Convenience method that calls feature.boruta.comp with preset parameters that are commonly used for learners in this package.

Usage

feature.boruta(data = data_train_numeric_clean_imputed, recompute = F,
  desc = "")

Arguments

data	data.frame containing the data on which the Boruta feature selection should be executed. Will be used to feed the target and predictors variables of feature.boruta.comp. Defaults to data_train_numeric_clean_imputed.
recompute	boolean switch that determines if the Boruta feature selection should be repeated when the results of a prior run have been found. Recommended when parameters have changed. Defaults to FALSE.
desc	Additional comment that can be appended to the name of the saved Boruta object. Can e.g. be used to store different Boruta runs for different learners. Defaults to the empty string.

Details

This method executes the packages Boruta wrapper feature.boruta.comp with all parameters preset to fit the needs of the package learners. Furthermore it stores each Boruta object after the computation to the output/feature_selection directory. Before a new computation is started the directory is checked for the existance of an already computed Boruta object. If one is available and the recompute flag is FALSE the previously computed object is loaded and used. Finally the method binds all selected features (confirmed and tentative ones) to the .GlobalEnv as the features_boruta variable.

Examples

 KaggleHouse:::feature_boruta()
 KaggleHouse:::feature_boruta(recompute = T)
 KaggleHouse:::feature_boruta(recompute = T, desc = "_test_run_")

MarcoNiemann/kaggle_house documentation built on May 7, 2019, 2:50 p.m.