feature.selection: Feature Selection.
In sbbrandt/xgbAnalysis: Reaction coordinates analysis

Description Usage Arguments

Dismisses features iteratively by their importance and monitor classification accuracy.
Trains iteratively a xgb model while dismissing features by importance and monitoring accuracy. Decreasing will dismiss the most important feature each round, increasing (default) the least important. The file 'feature.importance' will contain an ordered list of all dismissed features.
The parameter values used for feature selection are written to the file 'parameter'. Additionally, in savemode = T, the following files will be created each round i:

round<i>.model, where the trained model is saved.
round<i>.prediction, where the predictions are saved.
round<i>.importance, where the feature importanec is saved.

1
2
3

feature.selection(output_dir = "./featureSelection", data = "./data",
  decreasing = F, ndismiss = NA, fdimissed = NA, nrounds = 20,
  nthread = 0, eta = NA, max_depth = NA, savemode = T)