mlr3featsel adds filters, feature selection methods and embedded feature selection methods of algorithms to mlr3.

Travis build status CRAN status lifecycle Coverage status

Installation

remotes::install_github("mlr-org/mlr3featsel")

Filters

Public Methods

  • .$calculate(): Calculates Filter values
  • .$filter_*(): filters the task by a given criterion

  • .$scores: Filter score values
  • .$filtered_task: Filtered task

Generic Filters

Name Task Type Task Properties Param Set Feature Types Package
auc Classif twoclass Integer, Numeric Metrics
disr Classif character(0) Integer, Numeric, Factor, Ordered praznik
embedded Classif character(0) Logical, Integer, Numeric, Character, Factor, Ordered rpart
jmi Classif character(0) Integer, Numeric, Factor, Ordered praznik
kruskal_test Classif character(0) Integer, Numeric stats
mim Classif character(0) Integer, Numeric, Factor, Ordered praznik
njmim Classif character(0) Integer, Numeric, Factor, Ordered praznik
cmim Classif & Regr character(0) Integer, Numeric, Factor, Ordered praznik
gain_ratio Classif & Regr character(0) Integer, Numeric, Factor, Ordered FSelectorRcpp
information_gain Classif & Regr character(0) Integer, Numeric, Factor, Ordered FSelectorRcpp
symmetrical_uncertainty Classif & Regr character(0) Integer, Numeric, Factor, Ordered FSelectorRcpp
variance Classif & Regr character(0) Integer, Numeric stats
carscore Regr character(0) Numeric care
correlation Regr character(0) Integer, Numeric stats

Embedded Filters

The following learners have embedded filter methods which are supported via class FilterEmbedded:

## [1] "classif.featureless" "classif.ranger"      "classif.rpart"      
## [4] "classif.xgboost"     "regr.featureless"    "regr.ranger"        
## [7] "regr.rpart"          "regr.xgboost"

If your learner is listed here, the reason is most likely that it is not integrated into mlr3learners or mlr3extralearners. Please open an issue so we can add your package.

Some learners need to have their variable importance measure “activated” during learner creation. For example, to use the “impurity” measure of Random Forest via the ranger package:

##        score      feature   method
## 1: 44.498537  Petal.Width embedded
## 2: 42.552544 Petal.Length embedded
## 3:  9.878309 Sepal.Length embedded

“Wrapper” Methods

Work in progress.