Help for package mlr3filters

Title:

Filter Based Feature Selection for 'mlr3'

Version:

0.8.1

Description:

Extends 'mlr3' with filter methods for feature selection. Besides standalone filter methods built-in methods of any machine-learning algorithm are supported. Partial scoring of multivariate filter methods is supported.

License:

LGPL-3

URL:

https://mlr3filters.mlr-org.com, https://github.com/mlr-org/mlr3filters

BugReports:

https://github.com/mlr-org/mlr3filters/issues

Depends:

R (≥ 3.1.0)

Imports:

backports, checkmate, data.table, mlr3 (≥ 0.12.0), mlr3misc, paradox, R6

Suggests:

Boruta, care, caret, carSurv, FSelectorRcpp, knitr, lgr, mlr3learners, mlr3measures, mlr3pipelines, praznik, rpart, survival, testthat (≥ 3.0.0), withr

Config/testthat/edition:

Encoding:

UTF-8

NeedsCompilation:

RoxygenNote:

7.3.2

Collate:

'Filter.R' 'mlr_filters.R' 'FilterAUC.R' 'FilterAnova.R' 'FilterBoruta.R' 'FilterCMIM.R' 'FilterCarScore.R' 'FilterCarSurvScore.R' 'FilterCorrelation.R' 'FilterDISR.R' 'FilterFindCorrelation.R' 'FilterLearner.R' 'FilterImportance.R' 'FilterInformationGain.R' 'FilterJMI.R' 'FilterJMIM.R' 'FilterKruskalTest.R' 'FilterMIM.R' 'FilterMRMR.R' 'FilterNJMIM.R' 'FilterPerformance.R' 'FilterPermutation.R' 'FilterRelief.R' 'FilterSelectedFeatures.R' 'FilterUnivariateCox.R' 'FilterVariance.R' 'bibentries.R' 'flt.R' 'helper.R' 'reexports.R' 'zzz.R'

Packaged:

2024-11-08 10:12:13 UTC; marc

Author:

Marc Becker

[cre, aut], Patrick Schratz

[aut], Michel Lang

[aut], Bernd Bischl

[aut], Martin Binder [aut], John Zobolas

[aut]

Maintainer:

Marc Becker <marcbecker@posteo.de>

Repository:

CRAN

Date/Publication:

2024-11-08 11:30:02 UTC

mlr3filters: Filter Based Feature Selection for 'mlr3'

Description

Author(s)

Maintainer: Marc Becker marcbecker@posteo.de (ORCID)

Authors:

Patrick Schratz patrick.schratz@gmail.com (ORCID)
Michel Lang michellang@gmail.com (ORCID)
Bernd Bischl bernd_bischl@gmx.net (ORCID)
Martin Binder mlr.developer@mb706.com
John Zobolas bblodfon@gmail.com (ORCID)

Filter Base Class

Description

Base class for filters. Predefined filters are stored in the dictionary mlr_filters. A Filter calculates a score for each feature of a task. Important features get a large value and unimportant features get a small value. Note that filter scores may also be negative.

Details

Some features support partial scoring of the feature set: If nfeat is not NULL, only the best nfeat features are guaranteed to get a score. Additional features may be ignored for computational reasons, and then get a score value of NA.

Public fields

id

(character(1))
Identifier of the object. Used in tables, plot and text output.

label

(character(1))
Label for this object. Can be used in tables, plot and text output instead of the ID.

task_types

(character())
Set of supported task types, e.g. "classif" or "regr". Can be set to the scalar value NA to allow any task type.

For a complete list of possible task types (depending on the loaded packages), see mlr_reflections$task_types$type.

task_properties

(character())
mlr3::Tasktask properties.

param_set

(paradox::ParamSet)
Set of hyperparameters.

feature_types

(character())
Feature types of the filter.

packages

(character())
Packages which this filter is relying on.

man

(character(1))
String in the format ⁠[pkg]::[topic]⁠ pointing to a manual page for this object. Defaults to NA, but can be set by child classes.

scores

Stores the calculated filter score values as named numeric vector. The vector is sorted in decreasing order with possible NA values last. The more important the feature, the higher the score. Tied values (this includes NA values) appear in a random, non-deterministic order.

Active bindings

properties: (character())
Properties of the filter. Currently, only "missings" is supported. A filter has the property "missings", iff the filter can handle missing values in the features in a graceful way. Otherwise, an assertion is thrown if missing values are detected.
hash: (character(1))
Hash (unique identifier) for this object.
phash: (character(1))
Hash (unique identifier) for this partial object, excluding some components which are varied systematically during tuning (parameter values) or feature selection (feature names).

Methods

Method `new()`

Create a Filter object.

Usage

Filter$new(
  id,
  task_types,
  task_properties = character(),
  param_set = ps(),
  feature_types = character(),
  packages = character(),
  label = NA_character_,
  man = NA_character_
)

Arguments

id: (character(1))
Identifier for the filter.
task_types: (character())
Types of the task the filter can operator on. E.g., "classif" or "regr". Can be set to scalar NA to allow any task type.
task_properties: (character())
Required task properties, see mlr3::Task. Must be a subset of mlr_reflections$task_properties.
param_set: (paradox::ParamSet)
Set of hyperparameters.
feature_types: (character())
Feature types the filter operates on. Must be a subset of mlr_reflections$task_feature_types.
packages: (character())
Set of required packages. Note that these packages will be loaded via requireNamespace(), and are not attached.
label: (character(1))
Label for the new instance.
man: (character(1))
String in the format ⁠[pkg]::[topic]⁠ pointing to a manual page for this object. The referenced help package can be opened via method ⁠$help()⁠.

Method `format()`

Format helper for Filter class

Usage

Filter$format(...)

Arguments

...: (ignored).

Method `print()`

Printer for Filter class

Usage

Filter$print()

Method `help()`

Opens the corresponding help page referenced by field ⁠$man⁠.

Usage

Filter$help()

Method `calculate()`

Calculates the filter score values for the provided mlr3::Task and stores them in field scores. nfeat determines the minimum number of features to score (see details), and defaults to the number of features in task. Loads required packages and then calls private$.calculate() of the respective subclass.

This private method is is expected to return a numeric vector, uniquely named with (a subset of) feature names. The returned vector may have missing values. Features with missing values as well as features with no calculated score are automatically ranked last, in a random order. If the task has no rows, each feature gets the score NA.

Usage

Filter$calculate(task, nfeat = NULL)

Arguments

task: (mlr3::Task)
mlr3::Task to calculate the filter scores for.
nfeat: (integer())
The minimum number of features to calculate filter scores for.

Method `clone()`

The objects of this class are cloneable with this method.

Usage

Filter$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Syntactic Sugar for Filter Construction

Description

These functions complements mlr_filters with a function in the spirit of mlr3::mlr_sugar.

Usage

flt(.key, ...)

flts(.keys, ...)

Arguments

.key

(character(1))
Key passed to the respective dictionary to retrieve the object.

...

(any)
Additional arguments.

.keys

(character())
Keys passed to the respective dictionary to retrieve multiple objects.

Value

Filter.

Examples

flt("correlation", method = "kendall")
flts(c("mrmr", "jmim"))

Dictionary of Filters

Description

A simple mlr3misc::Dictionary storing objects of class Filter. Each Filter has an associated help page, see mlr_filters_[id].

This dictionary can get populated with additional filters by add-on packages.

For a more convenient way to retrieve and construct filters, see flt().

Usage

mlr_filters

Format

R6::R6Class object

Usage

See mlr3misc::Dictionary.

Examples

mlr_filters$keys()
as.data.table(mlr_filters)
mlr_filters$get("mim")
flt("anova")

ANOVA F-Test Filter

Description

ANOVA F-Test filter calling stats::aov(). Note that this is equivalent to a t-test for binary classification.

The filter value is -log10(p) where p is the p-value. This transformation is necessary to ensure numerical stability for very small p-values.

Super class

mlr3filters::Filter -> FilterAnova

Methods

Public methods

FilterAnova$new()
FilterAnova$clone()

Inherited methods

Method `new()`

Create a FilterAnova object.

Usage

FilterAnova$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterAnova$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

For a benchmark of filter methods:

Bommert A, Sun X, Bischl B, Rahnenführer J, Lang M (2020). “Benchmark for filter methods for feature selection in high-dimensional classification data.” Computational Statistics & Data Analysis, 143, 106839. doi:10.1016/j.csda.2019.106839.

Examples

task = mlr3::tsk("iris")
filter = flt("anova")
filter$calculate(task)
head(as.data.table(filter), 3)

# transform to p-value
10^(-filter$scores)

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("anova"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

AUC Filter

Description

Area under the (ROC) Curve filter, analogously to mlr3measures::auc() from mlr3measures. Missing values of the features are removed before calculating the AUC. If the AUC is undefined for the input, it is set to 0.5 (random classifier). The absolute value of the difference between the AUC and 0.5 is used as final filter value.

Super class

mlr3filters::Filter -> FilterAUC

Methods

Public methods

FilterAUC$new()
FilterAUC$clone()

Inherited methods

Method `new()`

Create a FilterAUC object.

Usage

FilterAUC$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterAUC$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

For a benchmark of filter methods:

Examples

task = mlr3::tsk("sonar")
filter = flt("auc")
filter$calculate(task)
head(as.data.table(filter), 3)

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("auc"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Burota Filter

Description

Filter using the Boruta algorithm for feature selection. If keep = "tentative", confirmed and tentative features are returned. Note that there is no ordering in the selected features. Selected features get a score of 1, deselected features get a score of 0. The order of selected features is random. In combination with mlr3pipelines, only the filter criterion cutoff makes sense.

Initial parameter values

num.threads:
- Actual default: NULL, triggering auto-detection of the number of CPUs.
- Adjusted value: 1.
- Reason for change: Conflicting with parallelization via future.

Super class

mlr3filters::Filter -> FilterBoruta

Methods

Public methods

FilterBoruta$new()
FilterBoruta$clone()

Inherited methods

Method `new()`

Creates a new instance of this R6 class.

Usage

FilterBoruta$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterBoruta$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

Kursa MB, Rudnicki WR (2010). “Feature Selection with the Boruta Package.” Journal of Statistical Software, 36(11), 1-13.

Examples


  if (requireNamespace("Boruta")) {
   task = mlr3::tsk("sonar")
   filter = flt("boruta")
   filter$calculate(task)
   as.data.table(filter)
  }

Correlation-Adjusted Marignal Correlation Score Filter

Description

Calculates the Correlation-Adjusted (marginal) coRrelation scores (short CAR scores) implemented in care::carscore() in package care. The CAR scores for a set of features are defined as the correlations between the target and the decorrelated features. The filter returns the absolute value of the calculated scores.

Argument verbose defaults to FALSE.

Super class

mlr3filters::Filter -> FilterCarScore

Methods

Public methods

FilterCarScore$new()
FilterCarScore$clone()

Inherited methods

Method `new()`

Create a FilterCarScore object.

Usage

FilterCarScore$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterCarScore$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (requireNamespace("care")) {
  task = mlr3::tsk("mtcars")
  filter = flt("carscore")
  filter$calculate(task)
  head(as.data.table(filter), 3)

  ## changing the filter settings
  filter = flt("carscore")
  filter$param_set$values = list("diagonal" = TRUE)
  filter$calculate(task)
  head(as.data.table(filter), 3)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "care", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("mtcars")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("carscore"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("regr.rpart"))

  graph$train(task)
}

Correlation-Adjusted Survival Score Filter

Description

Calculates CARS scores for right-censored survival tasks. Calls the implementation in carSurv::carSurvScore() in package carSurv.

Super class

mlr3filters::Filter -> FilterCarSurvScore

Methods

Public methods

FilterCarSurvScore$new()
FilterCarSurvScore$clone()

Inherited methods

Method `new()`

Create a FilterCarSurvScore object.

Usage

FilterCarSurvScore$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterCarSurvScore$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

Bommert A, Welchowski T, Schmid M, Rahnenführer J (2021). “Benchmark of filter methods for feature selection in high-dimensional gene expression survival data.” Briefings in Bioinformatics, 23(1). doi:10.1093/bib/bbab354.

Minimal Conditional Mutual Information Maximization Filter

Description

Minimal conditional mutual information maximization filter calling praznik::CMIM() from package praznik.

This filter supports partial scoring (see Filter).

Details

As the scores calculated by the praznik package are not monotone due to the greedy forward fashion, the returned scores simply reflect the selection order: 1, (k-1)/k, ..., 1/k where k is the number of selected features.

Threading is disabled by default (hyperparameter threads is set to 1). Set to a number ⁠>= 2⁠ to enable threading, or to 0 for auto-detecting the number of available cores.

Super class

mlr3filters::Filter -> FilterCMIM

Methods

Public methods

FilterCMIM$new()
FilterCMIM$clone()

Inherited methods

Method `new()`

Create a FilterCMIM object.

Usage

FilterCMIM$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterCMIM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

Kursa MB (2021). “Praznik: High performance information-based feature selection.” SoftwareX, 16, 100819. doi:10.1016/j.softx.2021.100819.

For a benchmark of filter methods:

Examples

if (requireNamespace("praznik")) {
  task = mlr3::tsk("iris")
  filter = flt("cmim")
  filter$calculate(task, nfeat = 2)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart", "praznik"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("cmim"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Correlation Filter

Description

Simple correlation filter calling stats::cor(). The filter score is the absolute value of the correlation.

Super class

mlr3filters::Filter -> FilterCorrelation

Methods

Public methods

FilterCorrelation$new()
FilterCorrelation$clone()

Inherited methods

Method `new()`

Create a FilterCorrelation object.

Usage

FilterCorrelation$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterCorrelation$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Note

This filter, in its default settings, can handle missing values in the features. However, the resulting filter scores may be misleading or at least difficult to compare if some features have a large proportion of missing values.

If a feature has no non-missing value, the resulting score will be NA. Missing scores appear in a random, non-deterministic order at the end of the vector of scores.

References

For a benchmark of filter methods:

Examples

## Pearson (default)
task = mlr3::tsk("mtcars")
filter = flt("correlation")
filter$calculate(task)
as.data.table(filter)

## Spearman
filter = FilterCorrelation$new()
filter$param_set$values = list("method" = "spearman")
filter$calculate(task)
as.data.table(filter)
if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("mtcars")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("correlation"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("regr.rpart"))

  graph$train(task)
}

Double Input Symmetrical Relevance Filter

Description

Double input symmetrical relevance filter calling praznik::DISR() from package praznik.

This filter supports partial scoring (see Filter).

Details

Threading is disabled by default (hyperparameter threads is set to 1). Set to a number ⁠>= 2⁠ to enable threading, or to 0 for auto-detecting the number of available cores.

Super class

mlr3filters::Filter -> FilterDISR

Methods

Public methods

FilterDISR$new()
FilterDISR$clone()

Inherited methods

Method `new()`

Create a FilterDISR object.

Usage

FilterDISR$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterDISR$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

Kursa MB (2021). “Praznik: High performance information-based feature selection.” SoftwareX, 16, 100819. doi:10.1016/j.softx.2021.100819.

For a benchmark of filter methods:

Examples

if (requireNamespace("praznik")) {
  task = mlr3::tsk("iris")
  filter = flt("disr")
  filter$calculate(task)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart", "praznik"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("disr"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Correlation Filter

Description

Simple filter emulating caret::findCorrelation(exact = FALSE).

This gives each feature a score between 0 and 1 that is one minus the cutoff value for which it is excluded when using caret::findCorrelation(). The negative is used because caret::findCorrelation() excludes everything above a cutoff, while filters exclude everything below a cutoff. Here the filter scores are shifted by +1 to get positive values for to align with the way other filters work.

Subsequently caret::findCorrelation(cutoff = 0.9) lists the same features that are excluded with FilterFindCorrelation at score 0.1 (= 1 - 0.9).

Super class

mlr3filters::Filter -> FilterFindCorrelation

Methods

Public methods

FilterFindCorrelation$new()
FilterFindCorrelation$clone()

Inherited methods

Method `new()`

Create a FilterFindCorrelation object.

Usage

FilterFindCorrelation$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterFindCorrelation$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

# Pearson (default)
task = mlr3::tsk("mtcars")
filter = flt("find_correlation")
filter$calculate(task)
as.data.table(filter)

## Spearman
filter = flt("find_correlation", method = "spearman")
filter$calculate(task)
as.data.table(filter)

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("find_correlation"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Filter for Embedded Feature Selection via Variable Importance

Description

Variable Importance filter using embedded feature selection of machine learning algorithms. Takes a mlr3::Learner which is capable of extracting the variable importance (property "importance"), fits the model and extracts the importance values to use as filter scores.

Super classes

mlr3filters::Filter -> mlr3filters::FilterLearner -> FilterImportance

Public fields

learner: (mlr3::Learner)
Learner to extract the importance values from.

Methods

Public methods

FilterImportance$new()
FilterImportance$clone()

Inherited methods

Method `new()`

Create a FilterImportance object.

Usage

FilterImportance$new(learner = mlr3::lrn("classif.featureless"))

Arguments

learner: (mlr3::Learner)
Learner to extract the importance values from.

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterImportance$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (requireNamespace("rpart")) {
  task = mlr3::tsk("iris")
  learner = mlr3::lrn("classif.rpart")
  filter = flt("importance", learner = learner)
  filter$calculate(task)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart", "mlr3learners"), quietly = TRUE)) {
  library("mlr3learners")
  library("mlr3pipelines")
  task = mlr3::tsk("sonar")

  learner = mlr3::lrn("classif.rpart")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("importance", learner = learner), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.log_reg"))

  graph$train(task)
}

Information Gain Filter

Description

Information gain filter calling FSelectorRcpp::information_gain() in package FSelectorRcpp. Set parameter "type" to "gainratio" to calculate the gain ratio, or set to "symuncert" to calculate the symmetrical uncertainty (see FSelectorRcpp::information_gain()). Default is "infogain".

Argument equal defaults to FALSE for classification tasks, and to TRUE for regression tasks.

Super class

mlr3filters::Filter -> FilterInformationGain

Methods

Public methods

FilterInformationGain$new()
FilterInformationGain$clone()

Inherited methods

Method `new()`

Create a FilterInformationGain object.

Usage

FilterInformationGain$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterInformationGain$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (requireNamespace("FSelectorRcpp")) {
  ## InfoGain (default)
  task = mlr3::tsk("sonar")
  filter = flt("information_gain")
  filter$calculate(task)
  head(filter$scores, 3)
  as.data.table(filter)

  ## GainRatio

  filterGR = flt("information_gain")
  filterGR$param_set$values = list("type" = "gainratio")
  filterGR$calculate(task)
  head(as.data.table(filterGR), 3)

}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "FSelectorRcpp", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("information_gain"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)

}

Joint Mutual Information Filter

Description

Joint mutual information filter calling praznik::JMI() in package praznik.

This filter supports partial scoring (see Filter).

Details

Threading is disabled by default (hyperparameter threads is set to 1). Set to a number ⁠>= 2⁠ to enable threading, or to 0 for auto-detecting the number of available cores.

Super class

mlr3filters::Filter -> FilterJMI

Methods

Public methods

FilterJMI$new()
FilterJMI$clone()

Inherited methods

Method `new()`

Create a FilterJMI object.

Usage

FilterJMI$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterJMI$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

Kursa MB (2021). “Praznik: High performance information-based feature selection.” SoftwareX, 16, 100819. doi:10.1016/j.softx.2021.100819.

For a benchmark of filter methods:

Examples

if (requireNamespace("praznik")) {
  task = mlr3::tsk("iris")
  filter = flt("jmi")
  filter$calculate(task, nfeat = 2)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart", "praznik"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("jmi"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Minimal Joint Mutual Information Maximization Filter

Description

Minimal joint mutual information maximization filter calling praznik::JMIM() in package praznik.

This filter supports partial scoring (see Filter).

Details

Threading is disabled by default (hyperparameter threads is set to 1). Set to a number ⁠>= 2⁠ to enable threading, or to 0 for auto-detecting the number of available cores.

Super class

mlr3filters::Filter -> FilterJMIM

Methods

Public methods

FilterJMIM$new()
FilterJMIM$clone()

Inherited methods

Method `new()`

Create a FilterJMIM object.

Usage

FilterJMIM$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterJMIM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

Kursa MB (2021). “Praznik: High performance information-based feature selection.” SoftwareX, 16, 100819. doi:10.1016/j.softx.2021.100819.

For a benchmark of filter methods:

Examples

if (requireNamespace("praznik")) {
  task = mlr3::tsk("iris")
  filter = flt("jmim")
  filter$calculate(task, nfeat = 2)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart", "praznik"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("jmim"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Kruskal-Wallis Test Filter

Description

Kruskal-Wallis rank sum test filter calling stats::kruskal.test().

The filter value is -log10(p) where p is the p-value. This transformation is necessary to ensure numerical stability for very small p-values.

Super class

mlr3filters::Filter -> FilterKruskalTest

Methods

Public methods

FilterKruskalTest$new()
FilterKruskalTest$clone()

Inherited methods

Method `new()`

Create a FilterKruskalTest object.

Usage

FilterKruskalTest$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterKruskalTest$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Note

If a feature has not at least one non-missing observation per label, the resulting score will be NA. Missing scores appear in a random, non-deterministic order at the end of the vector of scores.

References

For a benchmark of filter methods:

Examples

task = mlr3::tsk("iris")
filter = flt("kruskal_test")
filter$calculate(task)
as.data.table(filter)

# transform to p-value
10^(-filter$scores)

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("kruskal_test"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Mutual Information Maximization Filter

Description

Conditional mutual information based feature selection filter calling praznik::MIM() in package praznik.

This filter supports partial scoring (see Filter).

Details

Threading is disabled by default (hyperparameter threads is set to 1). Set to a number ⁠>= 2⁠ to enable threading, or to 0 for auto-detecting the number of available cores.

Super class

mlr3filters::Filter -> FilterMIM

Methods

Public methods

FilterMIM$new()
FilterMIM$clone()

Inherited methods

Method `new()`

Create a FilterMIM object.

Usage

FilterMIM$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterMIM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

Kursa MB (2021). “Praznik: High performance information-based feature selection.” SoftwareX, 16, 100819. doi:10.1016/j.softx.2021.100819.

For a benchmark of filter methods:

Examples

if (requireNamespace("praznik")) {
  task = mlr3::tsk("iris")
  filter = flt("mim")
  filter$calculate(task, nfeat = 2)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart", "praznik"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("mim"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Minimum Redundancy Maximal Relevancy Filter

Description

Minimum redundancy maximal relevancy filter calling praznik::MRMR() in package praznik.

This filter supports partial scoring (see Filter).

Details

Threading is disabled by default (hyperparameter threads is set to 1). Set to a number ⁠>= 2⁠ to enable threading, or to 0 for auto-detecting the number of available cores.

Super class

mlr3filters::Filter -> FilterMRMR

Methods

Public methods

FilterMRMR$new()
FilterMRMR$clone()

Inherited methods

Method `new()`

Create a FilterMRMR object.

Usage

FilterMRMR$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterMRMR$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

Kursa MB (2021). “Praznik: High performance information-based feature selection.” SoftwareX, 16, 100819. doi:10.1016/j.softx.2021.100819.

For a benchmark of filter methods:

Examples

if (requireNamespace("praznik")) {
  task = mlr3::tsk("iris")
  filter = flt("mrmr")
  filter$calculate(task, nfeat = 2)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart", "praznik"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("mrmr"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Minimal Normalised Joint Mutual Information Maximization Filter

Description

Minimal normalised joint mutual information maximization filter calling praznik::NJMIM() from package praznik.

This filter supports partial scoring (see Filter).

Details

Threading is disabled by default (hyperparameter threads is set to 1). Set to a number ⁠>= 2⁠ to enable threading, or to 0 for auto-detecting the number of available cores.

Super class

mlr3filters::Filter -> FilterNJMIM

Methods

Public methods

FilterNJMIM$new()
FilterNJMIM$clone()

Inherited methods

Method `new()`

Create a FilterNJMIM object.

Usage

FilterNJMIM$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterNJMIM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

Kursa MB (2021). “Praznik: High performance information-based feature selection.” SoftwareX, 16, 100819. doi:10.1016/j.softx.2021.100819.

For a benchmark of filter methods:

Examples

if (requireNamespace("praznik")) {
  task = mlr3::tsk("iris")
  filter = flt("njmim")
  filter$calculate(task, nfeat = 2)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart", "praznik"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("njmim"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Predictive Performance Filter

Description

Filter which uses the predictive performance of a mlr3::Learner as filter score. Performs a mlr3::resample() for each feature separately. The filter score is the aggregated performance of the mlr3::Measure, or the negated aggregated performance if the measure has to be minimized.

Super classes

mlr3filters::Filter -> mlr3filters::FilterLearner -> FilterPerformance

Public fields

learner: (mlr3::Learner)
resampling: (mlr3::Resampling)
measure: (mlr3::Measure)

Methods

Public methods

FilterPerformance$new()
FilterPerformance$clone()

Inherited methods

Method `new()`

Create a FilterDISR object.

Usage

FilterPerformance$new(
  learner = mlr3::lrn("classif.featureless"),
  resampling = mlr3::rsmp("holdout"),
  measure = NULL
)

Arguments

learner: (mlr3::Learner)
mlr3::Learner to use for model fitting.
resampling: (mlr3::Resampling)
mlr3::Resampling to be used within resampling.
measure: (mlr3::Measure)
mlr3::Measure to be used for evaluating the performance.

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterPerformance$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (requireNamespace("rpart")) {
  task = mlr3::tsk("iris")
  learner = mlr3::lrn("classif.rpart")
  filter = flt("performance", learner = learner)
  filter$calculate(task)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("iris")
  l = lrn("classif.rpart")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("performance", learner = l), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Permutation Score Filter

Description

The permutation filter randomly permutes the values of a single feature in a mlr3::Task to break the association with the response. The permuted feature, together with the unmodified features, is used to perform a mlr3::resample(). The permutation filter score is the difference between the aggregated performance of the mlr3::Measure and the performance estimated on the unmodified mlr3::Task.

Parameters

standardize: logical(1)
Standardize feature importance by maximum score.
nmc: integer(1)

Number of Monte-Carlo iterations to use in computing the feature importance.

Super classes

mlr3filters::Filter -> mlr3filters::FilterLearner -> FilterPermutation

Public fields

learner: (mlr3::Learner)
resampling: (mlr3::Resampling)
measure: (mlr3::Measure)

Active bindings

hash: (character(1))
Hash (unique identifier) for this object.
phash: (character(1))
Hash (unique identifier) for this partial object, excluding some components which are varied systematically during tuning (parameter values) or feature selection (feature names).

Methods

Public methods

FilterPermutation$new()
FilterPermutation$clone()

Inherited methods

Method `new()`

Create a FilterPermutation object.

Usage

FilterPermutation$new(
  learner = mlr3::lrn("classif.featureless"),
  resampling = mlr3::rsmp("holdout"),
  measure = NULL
)

Arguments

learner: (mlr3::Learner)
mlr3::Learner to use for model fitting.
resampling: (mlr3::Resampling)
mlr3::Resampling to be used within resampling.
measure: (mlr3::Measure)
mlr3::Measure to be used for evaluating the performance.

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterPermutation$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (requireNamespace("rpart")) {
  learner = mlr3::lrn("classif.rpart")
  resampling = mlr3::rsmp("holdout")
  measure = mlr3::msr("classif.acc")
  filter = flt("permutation", learner = learner, measure = measure, resampling = resampling,
    nmc = 2)
  task = mlr3::tsk("iris")
  filter$calculate(task)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("iris")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("permutation", nmc = 2), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

RELIEF Filter

Description

Information gain filter calling FSelectorRcpp::relief() in package FSelectorRcpp.

Super class

mlr3filters::Filter -> FilterRelief

Methods

Public methods

FilterRelief$new()
FilterRelief$clone()

Inherited methods

Method `new()`

Create a FilterRelief object.

Usage

FilterRelief$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterRelief$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Note

This filter can handle missing values in the features. However, the resulting filter scores may be misleading or at least difficult to compare if some features have a large proportion of missing values.

If a feature has no non-missing observation, the resulting score will be (close to) 0.

Examples

if (requireNamespace("FSelectorRcpp")) {
  ## Relief (default)
  task = mlr3::tsk("iris")
  filter = flt("relief")
  filter$calculate(task)
  head(filter$scores, 3)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "FSelectorRcpp", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("iris")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("relief"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Filter for Embedded Feature Selection

Description

Filter using embedded feature selection of machine learning algorithms. Takes a mlr3::Learner which is capable of extracting the selected features (property "selected_features"), fits the model and extracts the selected features.

Note that contrary to mlr_filters_importance, there is no ordering in the selected features. Selected features get a score of 1, deselected features get a score of 0. The order of selected features is random and different from the order in the learner. In combination with mlr3pipelines, only the filter criterion cutoff makes sense.

Super classes

mlr3filters::Filter -> mlr3filters::FilterLearner -> FilterSelectedFeatures

Public fields

learner: (mlr3::Learner)
Learner to extract the importance values from.

Methods

Public methods

FilterSelectedFeatures$new()
FilterSelectedFeatures$clone()

Inherited methods

Method `new()`

Create a FilterImportance object.

Usage

FilterSelectedFeatures$new(learner = mlr3::lrn("classif.featureless"))

Arguments

learner: (mlr3::Learner)
Learner to extract the selected features from.

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterSelectedFeatures$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (requireNamespace("rpart")) {
  task = mlr3::tsk("iris")
  learner = mlr3::lrn("classif.rpart")
  filter = flt("selected_features", learner = learner)
  filter$calculate(task)
  as.data.table(filter)
}

if (mlr3misc::require_namespaces(c("mlr3pipelines", "mlr3learners", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  library("mlr3learners")
  task = mlr3::tsk("sonar")

  filter = flt("selected_features", learner = lrn("classif.rpart"))

  # Note: All filter scores are either 0 or 1, i.e. setting `filter.cutoff = 0.5` means that
  # we select all "selected features".

  graph = po("filter", filter = filter, filter.cutoff = 0.5) %>>%
    po("learner", mlr3::lrn("classif.log_reg"))

  graph$train(task)
}

Univariate Cox Survival Filter

Description

Calculates scores for assessing the relationship between individual features and the time-to-event outcome (right-censored survival data) using a univariate Cox proportional hazards model. The goal is to determine which features have a statistically significant association with the event of interest, typically in the context of clinical or biomedical research.

This filter fits a Cox Proportional Hazards model using each feature independently and extracts the p-value that quantifies the significance of the feature's impact on survival. The filter value is -log10(p) where p is the p-value. This transformation is necessary to ensure numerical stability for very small p-values. Also higher values denote more important features. The filter works only for numeric features so please ensure that factor variables are properly encoded, e.g. using PipeOpEncode.

Super class

mlr3filters::Filter -> FilterUnivariateCox

Methods

Public methods

FilterUnivariateCox$new()
FilterUnivariateCox$clone()

Inherited methods

Method `new()`

Create a FilterUnivariateCox object.

Usage

FilterUnivariateCox$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterUnivariateCox$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


filter = flt("univariate_cox")
filter

Variance Filter

Description

Variance filter calling stats::var().

Argument na.rm defaults to TRUE here.

Super class

mlr3filters::Filter -> FilterVariance

Methods

Public methods

FilterVariance$new()
FilterVariance$clone()

Inherited methods

Method `new()`

Create a FilterVariance object.

Usage

FilterVariance$new()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

FilterVariance$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

References

For a benchmark of filter methods:

Examples

task = mlr3::tsk("mtcars")
filter = flt("variance")
filter$calculate(task)
head(filter$scores, 3)
as.data.table(filter)

if (mlr3misc::require_namespaces(c("mlr3pipelines", "rpart"), quietly = TRUE)) {
  library("mlr3pipelines")
  task = mlr3::tsk("spam")

  # Note: `filter.frac` is selected randomly and should be tuned.

  graph = po("filter", filter = flt("variance"), filter.frac = 0.5) %>>%
    po("learner", mlr3::lrn("classif.rpart"))

  graph$train(task)
}

Objects exported from other packages

Description

These objects are imported from other packages. Follow the links below to see their documentation.

data.table: as.data.table

mlr3filters: Filter Based Feature Selection for 'mlr3'

Description

Author(s)

See Also

Filter Base Class

Description

Details

Public fields

Active bindings

Methods

Public methods

Method new()

Usage

Arguments

Method format()

Usage

Arguments

Method print()

Usage

Method help()

Usage

Method calculate()

Usage

Arguments

Method clone()

Usage

Arguments

See Also

Syntactic Sugar for Filter Construction

Description

Usage

Arguments

Value

Examples

Dictionary of Filters

Description

Usage

Format

Usage

See Also

Examples

ANOVA F-Test Filter

Description

Super class

Methods

Public methods

Method new()

Usage

Method clone()

Usage

Arguments

References

See Also

Examples

AUC Filter

Description

Super class

Methods

Public methods

Method new()

Usage

Method clone()

Usage

Arguments

References

See Also

Examples

Burota Filter

Description

Initial parameter values

Super class

Methods

Public methods

Method new()

Usage

Method clone()

Usage

Arguments

References

See Also

Method `new()`

Method `format()`

Method `print()`

Method `help()`

Method `calculate()`

Method `clone()`

Method `new()`

Method `clone()`

Method `new()`

Method `clone()`

Method `new()`

Method `clone()`

Method `new()`

Method `clone()`

Method `new()`

Method `clone()`

Method `new()`

Method `clone()`

Method `new()`

Method `clone()`

Method `new()`

Method `clone()`