Merge pull request #122 from mlr-org/fix_lightgbm_categorical

fix categorical_features in lightgbm

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: mlr3extralearners
 Title: Extra Learners For mlr3
-Version: 0.5.12
+Version: 0.5.13
 Authors@R:
     c(person(given = "Raphael",
              family = "Sonabend",

NEWS.md

@@ -1,3 +1,7 @@
+# mlr3extralearners 0.5.13
+
+* Fix `categorical_features` in {lightgbm} learners
+
 # mlr3extralearners 0.5.12
 
 * Patch for `lightgbm` updates

R/learner_lightgbm_classif_lightgbm.R

@@ -19,6 +19,7 @@
 #' @details
 #' For categorical features either pre-process data by encoding columns or
 #' specify the categorical columns with the `categorical_feature` parameter.
+#' For this learner please do not prefix the categorical feature with `name:`.
 #'
 #' @template seealso_learner
 #' @template example
@@ -285,9 +286,12 @@ LearnerClassifLightGBM = R6Class("LearnerClassifLightGBM",
         dtrain = lightgbm::lgb.Dataset(
           data = as.matrix(task$data(rows = train_ids, cols = task$feature_names)),
           label = train_label,
-          free_raw_data = FALSE
+          free_raw_data = FALSE,
+          categorical_feature = pars$categorical_feature
         )
 
+        pars$categorical_feature <- NULL
+
         dtest = lightgbm::lgb.Dataset.create.valid(
           dataset = dtrain,
           data = as.matrix(task$data(rows = valid_ids, cols = task$feature_names)),
@@ -318,9 +322,12 @@ LearnerClassifLightGBM = R6Class("LearnerClassifLightGBM",
         dtrain = lightgbm::lgb.Dataset(
           data = as.matrix(task$data(cols = task$feature_names)),
           label = train_label,
-          free_raw_data = FALSE
+          free_raw_data = FALSE,
+          categorical_feature = pars$categorical_feature
         )
 
+        pars$categorical_feature <- NULL
+
         if ("weights" %in% task$properties) {
           dtrain$setinfo("weight", task$weights$weight)
         }

R/learner_lightgbm_regr_lightgbm.R

@@ -16,9 +16,9 @@
 #'   - Adjusted default: -1L
 #'   - Reason for change: Prevents accidental conflicts with mlr messaging system.
 #'
-#' @details
 #' For categorical features either pre-process data by encoding columns or
 #' specify the categorical columns with the `categorical_feature` parameter.
+#' For this learner please do not prefix the categorical feature with `name:`.
 #'
 #' @template seealso_learner
 #' @template example
@@ -233,9 +233,12 @@ LearnerRegrLightGBM = R6Class("LearnerRegrLightGBM",
         dtrain = lightgbm::lgb.Dataset(
           data = as.matrix(task$data(rows = train_ids, cols = task$feature_names)),
           label = as.matrix(task$data(rows = train_ids, cols = task$target_names)),
-          free_raw_data = FALSE
+          free_raw_data = FALSE,
+          categorical_feature = pars$categorical_feature
         )
 
+        pars$categorical_feature <- NULL
+
         valid_ids = task$row_roles$validation
         dtest = lightgbm::lgb.Dataset.create.valid(
           dataset = dtrain,
@@ -258,9 +261,12 @@ LearnerRegrLightGBM = R6Class("LearnerRegrLightGBM",
         dtrain = lightgbm::lgb.Dataset(
           data = as.matrix(task$data(cols = task$feature_names)),
           label = as.matrix(task$data(cols = task$target_names)),
-          free_raw_data = FALSE
+          free_raw_data = FALSE,
+          categorical_feature = pars$categorical_feature
         )
 
+        pars$categorical_feature <- NULL
+
         if ("weights" %in% task$properties) {
           dtrain$setinfo("weight", task$weights$weight)
         }