get_factor_split_levels: Get split criterion (levels) for factor or character...

In imbs-hl/timbR: Tree interpretation methods based on ranger

Get split criterion (levels) for factor or character variables Character variables are automatically turned into factors in function ranger() from the package "ranger" Considers the three possibilities to treat and split factors (Parameter "respect.unordered.factors" in ranger()): - Treat the factor like an ordinal variable ("none", "NULL" or "FALSE" in ranger()) - Treat the factor as nominal and check all 2-partitions for split ("partition") - Treat factor as ordinal and determine optimal order of levels for the random forest ("order", "TRUE" in ranger())

Description

Get split criterion (levels) for factor or character variables Character variables are automatically turned into factors in function ranger() from the package "ranger" Considers the three possibilities to treat and split factors (Parameter "respect.unordered.factors" in ranger()): - Treat the factor like an ordinal variable ("none", "NULL" or "FALSE" in ranger()) - Treat the factor as nominal and check all 2-partitions for split ("partition") - Treat factor as ordinal and determine optimal order of levels for the random forest ("order", "TRUE" in ranger())

Usage

get_factor_split_levels(
  node_id,
  split_variable,
  split_variable_type,
  split_value,
  train_data_df,
  rf_list
)

Arguments

node_id	Node ID of the node whose parent ID is to be determined
split_variable	Column name of the split variable in training data
split_variable_type	Type of split variable: "factor" for factors, "character" for character variable
split_value	Number(s) used for the split, for example the entry "splival" from the function treeInfo()
train_data_df	Data frame of the training data with which the random forest was trained
rf_list	Random forest, which is built like the one you get from ranger()

Value

Character with levels passed from the parent node to the considered node

Author(s)

Lea Louisa Kronziel, M.Sc.

imbs-hl/timbR documentation built on April 17, 2025, 2:08 p.m.