mlr3filters

mlr3filters adds filters, feature selection methods and embedded feature selection methods of algorithms to mlr3.

Installation

CRAN version

install.packages("mlr3filters")

Development version

remotes::install_github("mlr-org/mlr3filters")

Filters

Filter Example

library("mlr3")
library("mlr3filters")

task = tsk("pima")
filter = flt("auc")
as.data.table(filter$calculate(task))

##     feature     score
##      <char>     <num>
## 1:  glucose 0.2927906
## 2:  insulin 0.2316288
## 3:     mass 0.1870358
## 4:      age 0.1869403
## 5:  triceps 0.1625115
## 6: pregnant 0.1195149
## 7: pressure 0.1075760
## 8: pedigree 0.1062015

Implemented Filters

Variable Importance Filters

The following learners allow the extraction of variable importance and therefore are supported by FilterImportance:

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

If your learner is not listed here but capable of extracting variable importance from the fitted model, the reason is most likely that it is not yet integrated in the package mlr3learners or the extra learner organization. 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:

task = tsk("iris")
lrn = lrn("classif.ranger")
lrn$param_set$values = list(importance = "impurity")

filter = flt("importance", learner = lrn)
filter$calculate(task)
head(as.data.table(filter), 3)

##         feature     score
##          <char>     <num>
## 1:  Petal.Width 44.588117
## 2: Petal.Length 42.501367
## 3: Sepal.Length  9.898418

Performance Filter

FilterPerformance is a univariate filter method which calls resample() with every predictor variable in the dataset and ranks the final outcome using the supplied measure. Any learner can be passed to this filter with classif.rpart being the default. Of course, also regression learners can be passed if the task is of type “regr”.