Added support for feature_names: "auto" to Dataset

Author: ankane

CHANGELOG.md

@@ -1,6 +1,7 @@
 ## 0.4.0 (unreleased)
 
 - Added support for hashes to `predict` method
+- Added support for `feature_names: "auto"` to `Dataset`
 - Dropped support for Ruby < 3.1
 
 ## 0.3.4 (2024-07-28)

lib/lightgbm/dataset.rb

@@ -142,10 +142,18 @@ def construct
           ncol = data.column_count
           flat_data = data.to_a.flatten
         elsif daru?(data)
+          if @feature_names == "auto"
+            @feature_names = data.vectors.to_a
+          end
           nrow, ncol = data.shape
           flat_data = data.map_rows(&:to_a).flatten
-        elsif numo?(data) || rover?(data)
-          data = data.to_numo if rover?(data)
+        elsif numo?(data)
+          nrow, ncol = data.shape
+        elsif rover?(data)
+          if @feature_names == "auto"
+            @feature_names = data.keys
+          end
+          data = data.to_numo
           nrow, ncol = data.shape
         else
           nrow = data.count

test/dataset_test.rb

@@ -52,14 +52,19 @@ def test_dump_text
   def test_matrix
     data = Matrix.build(3, 3) { |row, col| row + col }
     label = Vector.elements([4, 5, 6])
-    LightGBM::Dataset.new(data, label: label)
+    dataset = LightGBM::Dataset.new(data, label: label)
+    assert_equal ["Column_0", "Column_1", "Column_2"], dataset.feature_names
   end
 
   def test_daru
     data = Daru::DataFrame.from_csv(data_path)
     label = data["y"]
     data = data.delete_vector("y")
-    LightGBM::Dataset.new(data, label: label)
+    dataset = LightGBM::Dataset.new(data, label: label)
+    assert_equal ["Column_0", "Column_1", "Column_2", "Column_3"], dataset.feature_names
+
+    dataset = LightGBM::Dataset.new(data, label: label, feature_names: "auto")
+    assert_equal ["x0", "x1", "x2", "x3"], dataset.feature_names
   end
 
   def test_numo
@@ -68,7 +73,8 @@ def test_numo
     require "numo/narray"
     data = Numo::DFloat.new(3, 5).seq
     label = Numo::DFloat.new(3).seq
-    LightGBM::Dataset.new(data, label: label)
+    dataset = LightGBM::Dataset.new(data, label: label)
+    assert_equal ["Column_0", "Column_1", "Column_2", "Column_3", "Column_4"], dataset.feature_names
   end
 
   def test_rover
@@ -77,7 +83,11 @@ def test_rover
     require "rover"
     data = Rover.read_csv(data_path)
     label = data.delete("y")
-    LightGBM::Dataset.new(data, label: label)
+    dataset = LightGBM::Dataset.new(data, label: label)
+    assert_equal ["Column_0", "Column_1", "Column_2", "Column_3"], dataset.feature_names
+
+    dataset = LightGBM::Dataset.new(data, label: label, feature_names: "auto")
+    assert_equal ["x0", "x1", "x2", "x3"], dataset.feature_names
   end
 
   def test_copy