learner.xgboost: Basic xgboost learner

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

Description

Parallel tuning function using xgboost that is either based on glmnet, boruta or the whole dataset.

Usage

learner.xgboost(data = data_train_numeric_clean_imputed, lasso = FALSE,
  boruta = FALSE)

Arguments

data	Input data which is default set to the numeric, imputed and cleaned training dataset
lasso	Boolean flag that shrinks data to the features of feature.lasso() stored in the variable features_lasso
boruta	Boolean flag that shrinks data to the features of boruta.lasso() stored in the variable features_boruta

Details

The default execution uses the whole training dataset. By setting either the lasso or boruta parameter to true the number of features is reduced according to the results of the feature selection. The number of rounds is set to 500 because including that parameter into the set of tuning parameters leads to worse results. The xgboost learner is wrapper in a Filter wrapper that uses chi squared as feature selection method. The result is three times cross validated at maximum 1000 experiments using irace as a control structure.

Value

0 as error output if both flags are set to true

Examples

KaggleHouse:::learner.xgboost(lasso=TRUE)

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