get_vimp_table-methods: Extract variable importance table.

In familiar: End-to-End Automated Machine Learning and Model Evaluation

Extract variable importance table.

Description

This method retrieves and parses variable importance tables from their respective vimpTable objects.

Usage

get_vimp_table(x, state = "ranked", ...)

## S4 method for signature 'list'
get_vimp_table(x, state = "ranked", ...)

## S4 method for signature 'character'
get_vimp_table(x, state = "ranked", ...)

## S4 method for signature 'vimpTable'
get_vimp_table(x, state = "ranked", ...)

## S4 method for signature 'NULL'
get_vimp_table(x, state = "ranked", ...)

## S4 method for signature 'experimentData'
get_vimp_table(x, state = "ranked", ...)

## S4 method for signature 'familiarModel'
get_vimp_table(x, state = "ranked", data = NULL, as_object = FALSE, ...)

Arguments

x	Variable importance (vimpTable) object, a list thereof, or one or more paths to these objects. This method extracts the variable importance table from such objects.
state	State of the returned variable importance table. This affects what contents are shown, and in which format. The variable importance table can be returned with the following states: initial: initial state, directly after the variable importance table is filled. The returned variable importance table shows the raw, un-processed data. decoded: depending on the variable importance method, the initial variable importance table may contain the scores of individual contrasts for categorical variables. When decoded, scores from all contrasts are aggregated to a single score for each feature. declustered: variable importance is determined from fully processed features, which includes clustering. This means that a single feature in the variable importance table may represent multiple original features. When a variable importance table has been declustered, all clusters have been turned into their constituent features. ranked (default): The scores have been used to create ranks, with lower ranks indicating better features. Internally, the variable importance table will go through each state, i.e. an variable importance table in the initial state will be decoded, declustered and then ranked prior to returning the variable importance table.
...	Unused arguments.
data	Internally used argument for use with familiarModel objects.
as_object	Internally used argument for use with familiarModel objects.

Value

A data.table with variable importance scores and, with state="ranked", the respective ranks.

familiar documentation built on Sept. 30, 2024, 9:18 a.m.