SMOTEBoost.BEM: SMOTEBoost variant of BEMBoost

In nunompmoniz/ReBoost: Boosting by Resampling Approaches for Imbalanced Numerical Prediction Tasks

Description

SMOTEBoost variant of BEMBoost

Usage

SMOTEBoost.BEM(
  form,
  train,
  test,
  t_final = 100,
  BEM = 0.5,
  perc.O = 1.5,
  rel.thr = 0.9,
  k = 3,
  coef = 1.5,
  ...
)

Arguments

form	The model formula.
train	A data.frame with the training data.
test	A data.frame with the test data.
t_final	The number of maximum boosting iterations. Default is 100.
BEM	Biggest error margin admissible. Defaults to 0.5.
perc.O	Percentage for Oversampling via SMOTE, i.e. percentage of extreme cases to be generated. Default is 1.5.
rel.thr	Relevance threshold. Default is 0.9.
k	Number of neighbours used in SMOTE. Defaults to 3.
coef	Coefficient used in boxplot statistics, which is used to create the relevance function. Default is 1.5.
...	Dots are passed to rpart

Value

Returns a vector with the predictions made by SMOTEBoost.BEM.

Examples

data(Boston,package="MASS")

idx <- sample(1:nrow(Boston),nrow(Boston)*0.75)
form <- medv ~ .

train <- Boston[idx,]
test <- Boston[-idx,]

preds <- SMOTEBoost.BEM(form,train,test)

nunompmoniz/ReBoost documentation built on Sept. 15, 2021, 8:28 p.m.