fi_methods: Feature Importance methods

In dynverse/dynfeature: Feature Importance for Dynamic Processes

Feature Importance methods

Description

Feature Importance methods

Usage

fi_ranger_rf_lite(
  num_trees = 2000,
  num_variables_per_split = 50,
  num_samples_per_tree = 250,
  min_node_size = 20,
  ...
)

fi_ranger_rf(...)

fi_caret(caret_method, ...)

fi_ranger_rf_tiny(
  num_trees = 100,
  num_variables_per_split = 50,
  num_samples_per_tree = 250,
  min_node_size = 20,
  ...
)

Arguments

num_trees	(fi_ranger_rf_lite) The number of trees to use
num_variables_per_split	(fi_ranger_rf_lite) The number of variables to sample per split
num_samples_per_tree	(fi_ranger_rf_lite) The number of samples to bootstrap per split
min_node_size	(fi_ranger_rf_lite) The minimum node size, no split will be made if the node size is less than this value.
...	Extra parameters to pass onto the underlying feature importance function.
caret_method	(fi_caret) Which caret method to use for feature importance.

Value

A list containing a helper function for calling a feature importance function.

Examples

library(dynwrap)
data(example_trajectory)

calculate_overall_feature_importance(example_trajectory, fi_method = fi_ranger_rf())

dynverse/dynfeature documentation built on March 3, 2023, 1:44 p.m.