feat: add feature_ranking method

R/EnsembleFSResult.R

@@ -45,9 +45,26 @@ EnsembleFSResult = R6Class("EnsembleFSResult",
     },
 
     #' @description
-    #' Returns the feature ranking.
-    feature_ranking = function() {
+    #' Calculates the feature ranking.
+    #'
+    #' @param method (`character(1)`)\cr
+    #' The method to calculate the feature ranking.
+    #' Currently, only `"inclusion_probability"` is supported.
+    feature_ranking = function(method = "inclusion_probability") {
+      assert_choice(method, choices = "inclusion_probability")
+
+      features = self$benchmark_result$tasks$task[[1]]$feature_names
+
+      count = map_int(features, function(feature) {
+        sum(map_lgl(self$result$features, function(iteration) {
+          feature %in% iteration
+        }))
+      })
+
+      res = data.table(feature = features, inclusion_probability = count / nrow(self$result))
+      setorderv(res, "inclusion_probability", order = -1L)
 
+      res
     },
 
     #' @description

tests/testthat/test_ensemble_fselect.R

@@ -15,20 +15,9 @@ test_that("ensemble feature selection works", {
   expect_vector(efsr$result$classif.ce, size = 4)
   expect_list(efsr$result$importance, any.missing = FALSE, len = 4)
   expect_benchmark_result(efsr$benchmark_result)
-})
-
-test_that("stability method works", {
-  efsr = ensemble_fselect(
-    fselector = fs("rfe", n_features = 2, feature_fraction = 0.8),
-    task = tsk("sonar"),
-    learners = lrns(c("classif.rpart", "classif.featureless")),
-    init_resampling = rsmp("subsampling", repeats = 2),
-    inner_resampling = rsmp("cv", folds = 3),
-    measure = msr("classif.ce"),
-    terminator = trm("none")
-  )
 
   expect_number(efsr$stability(stability_measure = "jaccard"))
+  feature_ranking = efsr$feature_ranking()
+  expect_data_table(feature_ranking, nrows = 60)
+  expect_names(names(feature_ranking), identical.to = c("feature", "inclusion_probability"))
 })
-
-