Improved handling of new and invalid categories

Author: ankane

lib/lightgbm/inner_predictor.rb

@@ -141,10 +141,17 @@ def cached_feature_name
     def apply_pandas_categorical(data, categorical_feature, pandas_categorical)
       (categorical_feature || []).each_with_index do |cf, i|
         cat_codes = pandas_categorical[i].map.with_index.to_h
-        # TODO confirm column is categorical
         data.each do |r|
-          # TODO decide how to handle missing values
-          r[cf] = cat_codes.fetch(r[cf])
+          cat = r[cf]
+          unless cat.nil?
+            r[cf] =
+              cat_codes.fetch(cat) do
+                unless cat.is_a?(String)
+                  raise ArgumentError, "expected categorical value"
+                end
+                nil
+              end
+          end
         end
       end
     end

test/booster_test.rb

@@ -144,6 +144,24 @@ def test_predict_pandas_categorical_model_str
     assert_elements_in_delta [0.996415541144579, 1.0809369939979934], y_pred.first(2)
   end
 
+  def test_predict_pandas_categorical_missing_category
+    booster = LightGBM::Booster.new(model_file: "test/support/categorical.txt")
+    assert_in_delta 0.996415541144579, booster.predict([3.7, 1.2, 7.2, nil])
+  end
+
+  def test_predict_pandas_categorical_new_category
+    booster = LightGBM::Booster.new(model_file: "test/support/categorical.txt")
+    assert_in_delta 0.996415541144579, booster.predict([3.7, 1.2, 7.2, "cat10"])
+  end
+
+  def test_predict_pandas_categorical_invalid_category
+    booster = LightGBM::Booster.new(model_file: "test/support/categorical.txt")
+    error = assert_raises(ArgumentError) do
+      booster.predict([7.5, 0.5, 7.9, true])
+    end
+    assert_equal "expected categorical value", error.message
+  end
+
   def test_model_to_string
     assert booster.model_to_string
   end