logicDT.boosting {logicDT}

Fitting boosted logicDT models

Description

Function for fitting gradient boosted logicDT models.

Usage

## Default S3 method:
logicDT.boosting(
  X,
  y,
  Z = NULL,
  boosting.iter = 500,
  learning.rate = 0.01,
  subsample.frac = 1,
  replace = TRUE,
  line.search = "min",
  ...
)

## S3 method for class 'formula'
logicDT.boosting(formula, data, ...)

Arguments

X

Matrix or data frame of binary predictors coded as 0 or 1.

y

Response vector. 0-1 coding for binary responses. Otherwise, a regression task is assumed.

Z

Optional matrix or data frame of quantitative/continuous covariables. Multiple covariables allowed for splitting the trees. If leaf regression models (such as four parameter logistic models) shall be fitted, only the first given covariable is used.

boosting.iter

Number of boosting iterations

learning.rate

Learning rate for boosted models. Values between 0.001 and 0.1 are recommended.

subsample.frac

Subsample fraction for each boosting iteration. E.g., 0.5 means that are random draw of 50 is used in each iteration.

replace

Should the random draws with subsample.frac in boosted models be performed with or without replacement? TRUE or FALSE

line.search

Type of line search for gradient boosting. "min" performs a real minimization while "binary" performs a loose binary search for a boosting coefficient that just reduces the score.

...

Arguments passed to logicDT

formula

An object of type formula describing the model to be fitted.

data

A data frame containing the data for the corresponding formula object. Must also contain quantitative covariables if they should be included as well.

Details

Details on single logicDT models can be found in logicDT.

Value

An object of class logic.boosted. This is a list containing

models

A list of fitted logicDT models

rho

A vector of boosting coefficient corresponding to each model

initialModel

Initial model which is usually the observed mean

...

Supplied parameters of the functional call to logicDT.boosting.

References

• Lau, M., Schikowski, T. & Schwender, H. (2024). logicDT: A procedure for identifying response-associated interactions between binary predictors. Machine Learning 113(2):933–992. doi: 10.1007/s10994-023-06488-6

• Friedman, J. H. (2001). Greedy Function Approximation: A Gradient Boosting Machine. The Annals of Statistics, 29(5), 1189–1232. doi: 10.1214/aos/1013203451