make_xgb_model: Make predictive models of dependencies

View source: R/make_xgb_models.R

make_xgb_modelR Documentation

Make predictive models of dependencies

Description

This function creates an XGBoost model

Usage

make_xgb_model(
  perturbation,
  indx,
  total,
  dataset,
  response_cutoff = 0.75,
  weight_cap = 0.05,
  nfolds = 3,
  nrepeats = 3,
  nrounds = 100,
  max_depth = 3,
  f_subsample = 1,
  min_score = 0.5,
  skip_eval = FALSE,
  shuffle = FALSE,
  n_threads = 4,
  xgb_params = NULL,
  use_gpu = TRUE,
  gpu_id = 0
)

Arguments

perturbation

Column name of the perturbation (e.g. "ko_ctnnb1").

indx

Integer index used, for progress report.

total

Integer of the total number of perturbations passed to this function, for progress report.

dataset

A dataframe with the perturbation in a column and all other predictors. Sample names are row names.

response_cutoff

The value above which the sample is considered sensitive.

weight_cap

The maximum weight of each minority case when resampling. Set to 0 if no resampling needed.

nfolds

The number of folds in k-fold cross validation.

nrepeats

The number of repeats in k-fold cross validation.

nrounds

The maximum number of trees in the XGBoost model.

min_score

The minimum number of r value for a model to be considered for the next stage (making predictions and calculating SHAP values).

skip_eval

Default = FALSE. If TRUE, k-fold CV will not be conducted and instead all models will be pushed to the next stage.

use_gpu

Default = TRUE. Set to FALSE if using CPU.

Examples

make_xgb_model("ko_ctnnb1",1,1,my_data)

Mushriq/mixmap documentation built on Jan. 28, 2024, 7:22 p.m.