Feature Selection via Sequential Selection

FSelectorSequential class that implements sequential feature selection. The sequential forward selection (strategy = fsf) extends the feature set in each step with the feature that increases the models performance the most. The sequential backward selection (strategy = fsb) starts with the complete future set and removes in each step the feature that decreases the models performance the least.

Note

Feature sets are evaluated in batches, where each batch is one step in the sequential feature selection.

Dictionary

This FSelector can be instantiated via the dictionary mlr_fselectors or with the associated sugar function fs():

mlr_fselectors$get("sequential")
fs("sequential")

Parameters

max_features

integer(1)
Maximum number of features. By default, number of features in mlr3::Task.

strategy

character(1)
Search method sfs (forward search) or sbs (backward search).

Super class

mlr3fselect::FSelector -> FSelectorSequential

Methods

Public methods

• FSelectorSequential$new()

• FSelectorSequential$optimization_path()

• FSelectorSequential$clone()

Inherited methods

• mlr3fselect::FSelector$format()

• mlr3fselect::FSelector$optimize()

• mlr3fselect::FSelector$print()

---

Method new()

Creates a new instance of this R6 class.`

Usage

FSelectorSequential$new()

---

Method optimization_path()

Returns the optimization path.

Usage

FSelectorSequential$optimization_path(inst, include_uhash = FALSE)

Arguments

inst

(FSelectInstanceSingleCrit)
Instance optimized with FSelectorSequential.

include_uhash

(logical(1))
Include uhash column?

Returns

data.table::data.table()

---

Method clone()

The objects of this class are cloneable with this method.

Usage

FSelectorSequential$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

library(mlr3)

terminator = trm("evals", n_evals = 5)

instance = FSelectInstanceSingleCrit$new(
  task = tsk("iris"),
  learner = lrn("classif.rpart"),
  resampling = rsmp("holdout"),
  measure = msr("classif.ce"),
  terminator = terminator
)

fselector = fs("sequential")
# \donttest{
# Modifies the instance by reference
fselector$optimize(instance)
#>    Petal.Length Petal.Width Sepal.Length Sepal.Width    features classif.ce
#> 1:        FALSE        TRUE        FALSE       FALSE Petal.Width       0.02

# Returns best scoring evaluation
instance$result
#>    Petal.Length Petal.Width Sepal.Length Sepal.Width    features classif.ce
#> 1:        FALSE        TRUE        FALSE       FALSE Petal.Width       0.02

# Allows access of data.table of full path of all evaluations
as.data.table(instance$archive)# }
#>    Petal.Length Petal.Width Sepal.Length Sepal.Width classif.ce
#> 1:         TRUE       FALSE        FALSE       FALSE       0.04
#> 2:        FALSE        TRUE        FALSE       FALSE       0.02
#> 3:        FALSE       FALSE         TRUE       FALSE       0.32
#> 4:        FALSE       FALSE        FALSE        TRUE       0.60
#> 5:         TRUE        TRUE        FALSE       FALSE       0.04
#> 6:        FALSE        TRUE         TRUE       FALSE       0.02
#> 7:        FALSE        TRUE        FALSE        TRUE       0.02
#>    runtime_learners           timestamp batch_nr      resample_result
#> 1:            0.064 2021-09-17 04:15:58        1 <ResampleResult[20]>
#> 2:            0.077 2021-09-17 04:15:58        1 <ResampleResult[20]>
#> 3:            0.076 2021-09-17 04:15:58        1 <ResampleResult[20]>
#> 4:            0.063 2021-09-17 04:15:58        1 <ResampleResult[20]>
#> 5:            0.067 2021-09-17 04:15:59        2 <ResampleResult[20]>
#> 6:            0.062 2021-09-17 04:15:59        2 <ResampleResult[20]>
#> 7:            0.063 2021-09-17 04:15:59        2 <ResampleResult[20]>