Skip to contents

This function performs feature selection using a logistic regression model and bootstrapping for each cluster in the dataset. For each cluster, the function evaluates which features are significantly associated with the cluster based on a logistic regression model with bootstrap sampling. The output includes a count of significant features for each cluster based on the p-value threshold.

Usage

FeatureSelectionWithBootstrap(
  data,
  p.no_bootstrap = 0.01,
  p.bootstrap = 0.05,
  num.iteration = 1000,
  nCores = parallel::detectCores() - 3
)

Arguments

data

A data frame where the first column is the cluster variable (categorical), and the other columns are feature values.

p.no_bootstrap

Numeric. The p-value threshold for the features to be selected based on the original data (default: 0.01).

p.bootstrap

Numeric. The p-value threshold for the features to be selected based on bootstrapped samples (default: 0.05).

num.iteration

Integer. The number of bootstrap iterations (default: 1000).

nCores

Integer. The number of CPU cores to use for parallel processing (default: automatically set to all cores minus 3).

Value

A list of data frames, each containing the features selected for each cluster and the count of significant features based on bootstrapping.

Author

Zaoqu Liu; Email: liuzaoqu@163.com