tidy_tree: Turn a classification/regression tree object into a tidy...

In bakaburg1/tidytrees: Convert Regression/Classification Trees Into Tidy Tibbles

Description

Takes a classification/regression tree (usually a list) and returns a tibble::tibble() with a row for each node including the set of additive rules necessary to identify it. Furthermore, node characteristics and fit details are described. Check https://github.com/bakaburg1/tidytrees for a more detailed explanation.

Usage

## S3 method for class 'party'
tidy_tree(
  tree,
  rule_as_text = TRUE,
  eval_ready = FALSE,
  simplify_rules = FALSE,
  add_estimates = TRUE,
  add_interval = FALSE,
  interval_level = 0.95,
  est_fun = tidytrees::get_pred_estimates
)

## S3 method for class 'rpart'
tidy_tree(
  tree,
  rule_as_text = TRUE,
  eval_ready = FALSE,
  simplify_rules = FALSE,
  add_estimates = TRUE,
  add_interval = FALSE,
  interval_level = 0.95,
  est_fun = tidytrees::get_pred_estimates
)

tidy_tree(
  tree,
  rule_as_text = TRUE,
  eval_ready = FALSE,
  simplify_rules = FALSE,
  add_estimates = TRUE,
  add_interval = FALSE,
  interval_level = 0.95,
  est_fun = tidytrees::get_pred_estimates
)

Arguments

tree	A tree object.
rule_as_text	Whether to represent the rules as a string or a vector.
eval_ready	Converts the rules into R compatible logical expressions ready to use for data filtering purposes. If FALSE, the rules are kept as originally defined in each class.
simplify_rules	Keep the minimal set of conditions to identify a node in the tree.
add_estimates	Add predicted values at each node, as computed by the function passed to est_fun.
add_interval	Logical indicating whether or not to include an estimation interval, as computed by est_fun in the tidied output. Defaults to FALSE.
interval_level	The interval level to use for the estimation interval if add_interval = TRUE. Must be strictly greater than 0 and less than 1. Defaults to 0.95.
est_fun	Function to estimate node predictions and intervals. Must expose three mandatory arguments: values which receive the observations in a node, and add_interval and interval_level which get inherited from tidy_tree(). Check get_pred_estimates as a prototype.
...	Method specific arguments. Not used at the moment.

Value

A tibble with a row for each node with its identifying rule, the node id (as stored in the tree object), the number of observations related to the node, whether the node is terminal (a leaf), and the node depth. The depth is counted starting from the children of the root node which are considered at depth 1. The root node is ignored in the output. If add_estimates = TRUE, prediction estimates (optionally with intervals) are added to each node, as defined by the function passed to the est_fun argument.

Methods (by class)

• party: Turns a classification/regression tree produced by partykit::ctree() into a tidy tibble. Also trees of class XML::XMLNode and RWeka::Weka_tree can be processed after conversion with partykit::as.party() (but results have not been tested yet).

• rpart: Turn a classification/regression tree produced by rpart::rpart() into a tidy tibble. At the moment, only the anova and class methods are fully supported: other methods may work but users need to ensure that the fit details are correct.

Examples

tree <- rpart(iris$Sepal.Length ~ iris$Species)

tidy_tree(mod)

## Adding confidence intervals

tidy_tree(mod, add_interval = TRUE)

## Simplify rules to remove redundant conditions

tree <- ctree(Sepal.Length ~ Species + Sepal.Width, iris)
tidy_tree(simplify_rules = TRUE)

bakaburg1/tidytrees documentation built on Dec. 19, 2021, 6:40 a.m.