sample_mle: Generate artifical data and train prediction models

In maxwestphal/SEPM.SIM: Statistical Evaluation of Prediction Models - Simulation Studies

Generate artifical data and train prediction models

Description

This function is used to generate artifical observations of feature-label pairs which are then used to train prediction models by various machine learning algorithms.

Usage

sample_mle(
  n.learn = 400,
  ratio.lv = 0.75,
  n.eval = 10000,
  n.pop = 1e+05,
  P = 50,
  red = 0,
  rho = 0.5,
  scenario = "EOMPM_A2",
  scenario.learn = scenario,
  methods = "glmnet_xgbTree_rpartCost_svmLinearWeights2",
  M = 50,
  tuning = "random",
  data = NULL,
  job = NULL
)

Arguments

n.learn	integer, number of learning observations
ratio.lv	numeric, fraction of learning observations to use for validation (default: 0.75)
n.eval	integer, number of test observations
n.pop	integer, population sample size (used to calculate 'true' performances)
P	integer, total number of features (needs to be larger than active features of scenario)
red	integer, set degree of redundancy among features (independent features with red=0)
rho	numeric, (equi)correlation in (0,1) between redundant features (no effect if red=0)
scenario	character, specifies data distribution scenario, e.g. "EOMPM_A2" (default)
scenario.learn	character, potentially perturbed learning distribution, equalls scenario by default
methods	character, specify learning algorithms and combinations thereof, e.g. "glmnet_xgbTree", the following methods are supported: "glmnet", "xgbTree", "rpartCost", "svmLinearWeights2" (see caret documentation; default: "glmnet_xgbTree_rpartCost_svmLinearWeights2")
M	integer, number of models to train per learning algorithm (default: 50)
tuning	character, specify hyperparameter sampling, either "random" (default) or "grid"
data	NULL, batchtools argument
job	NULL, batchtools argument

Value

Returns a data instance which is processed by either the study_mle_acc() or theory().

maxwestphal/SEPM.SIM documentation built on April 11, 2024, 4:06 p.m.