Added the 'input_float' tree model transformation option

jpmml · Mar 28, 2024 · 7e1292b · 7e1292b
1 parent 5d07e83
commit 7e1292b
Show file tree

Hide file tree

Showing 6 changed files with 82 additions and 6 deletions.
diff --git a/pmml-sklearn-example/src/main/java/org/jpmml/sklearn/example/Main.java b/pmml-sklearn-example/src/main/java/org/jpmml/sklearn/example/Main.java
@@ -85,6 +85,13 @@ public class Main {
 	)
 	private Boolean flat = null;
 
+	@Parameter (
+		names = {"--X-" + HasTreeOptions.OPTION_INPUT_FLOAT},
+		description = "Allow field data type updates",
+		arity = 1
+	)
+	private Boolean inputFloat = null;
+
 	@Parameter (
 		names = {"--X-" + HasTreeOptions.OPTION_NODE_ID},
 		description = "Keep SkLearn node identifiers",
@@ -219,6 +226,7 @@ public void run() throws Exception {
 		options.put(HasTreeOptions.OPTION_ALLOW_MISSING, this.allowMissing);
 		options.put(HasTreeOptions.OPTION_COMPACT, this.compact);
 		options.put(HasTreeOptions.OPTION_FLAT, this.flat);
+		options.put(HasTreeOptions.OPTION_INPUT_FLOAT, this.inputFloat);
 		options.put(HasTreeOptions.OPTION_NODE_ID, this.nodeId);
 		options.put(HasTreeOptions.OPTION_NODE_SCORE, this.nodeScore);
 		options.put(HasTreeOptions.OPTION_NUMERIC, this.numeric);

diff --git a/pmml-sklearn/src/main/java/sklearn/tree/HasTreeOptions.java b/pmml-sklearn/src/main/java/sklearn/tree/HasTreeOptions.java
@@ -21,6 +21,7 @@
 import java.util.LinkedHashMap;
 import java.util.Map;
 
+import org.dmg.pmml.DerivedField;
 import org.dmg.pmml.OutputField;
 import org.dmg.pmml.SimplePredicate;
 import org.dmg.pmml.SimpleSetPredicate;
@@ -48,6 +49,11 @@ public interface HasTreeOptions extends HasSkLearnOptions, HasNativeConfiguratio
 	 */
 	String OPTION_FLAT = "flat";
 
+	/**
+	 * @see DerivedField
+	 */
+	String OPTION_INPUT_FLOAT = "input_float";
+
 	/**
 	 * @see Node#hasExtensions()
 	 * @see Node#getExtensions()

diff --git a/pmml-sklearn/src/main/java/sklearn/tree/TreeUtil.java b/pmml-sklearn/src/main/java/sklearn/tree/TreeUtil.java
@@ -30,7 +30,10 @@
 import com.google.common.primitives.Doubles;
 import numpy.core.ScalarUtil;
 import org.dmg.pmml.DataType;
+import org.dmg.pmml.Field;
+import org.dmg.pmml.HasContinuousDomain;
 import org.dmg.pmml.HasExtensions;
+import org.dmg.pmml.Interval;
 import org.dmg.pmml.MiningFunction;
 import org.dmg.pmml.Model;
 import org.dmg.pmml.Predicate;
@@ -345,8 +348,9 @@ public Number get(int index){
 	static
 	public <E extends Estimator & HasTreeOptions> Schema configureSchema(E estimator, Schema schema){
 		Boolean numeric = (Boolean)estimator.getOption(HasTreeOptions.OPTION_NUMERIC, Boolean.TRUE);
+		Boolean inputFloat = (Boolean)estimator.getOption(HasTreeOptions.OPTION_INPUT_FLOAT, null);
 
-		return toTreeModelSchema(estimator.getDataType(), numeric, schema);
+		return toTreeModelSchema(numeric, inputFloat, schema);
 	}
 
 	static
@@ -567,7 +571,7 @@ private void encodeNodeId(Estimator estimator, Model model){
 	}
 
 	static
-	private Schema toTreeModelSchema(DataType dataType, boolean numeric, Schema schema){
+	private Schema toTreeModelSchema(Boolean numeric, Boolean inputFloat, Schema schema){
 		Function<Feature, Feature> function = new Function<Feature, Feature>(){
 
 			@Override
@@ -585,14 +589,51 @@ public Feature apply(Feature feature){
 					return missingValueFeature;
 				} else
 
-				if(feature instanceof ThresholdFeature && !numeric){
+				if(feature instanceof ThresholdFeature && (numeric != null && !numeric)){
 					ThresholdFeature thresholdFeature = (ThresholdFeature)feature;
 
 					return thresholdFeature;
 				} else
 
+				if(inputFloat != null && inputFloat){
+					ContinuousFeature continuousFeature = feature.toContinuousFeature();
+
+					DataType dataType = continuousFeature.getDataType();
+					if(dataType != DataType.FLOAT){
+						Field<?> field = continuousFeature.getField();
+
+						field.setDataType(DataType.FLOAT);
+
+						// XXX
+						if(field instanceof HasContinuousDomain){
+							HasContinuousDomain<?> hasContinuousDomain = (HasContinuousDomain<?>)field;
+
+							if(hasContinuousDomain.hasIntervals()){
+								List<Interval> intervals = hasContinuousDomain.getIntervals();
+
+								for(Interval interval : intervals){
+									Number leftMargin = interval.getLeftMargin();
+									Number rightMargin = interval.getRightMargin();
+
+									if(leftMargin != null){
+										interval.setLeftMargin((double)leftMargin.floatValue());
+									} // End if
+
+									if(rightMargin != null){
+										interval.setRightMargin((double)rightMargin.floatValue());
+									}
+								}
+							}
+						}
+
+						return new ContinuousFeature(continuousFeature.getEncoder(), field);
+					}
+
+					return continuousFeature;
+				} else
+
 				{
-					ContinuousFeature continuousFeature = feature.toContinuousFeature(dataType);
+					ContinuousFeature continuousFeature = feature.toContinuousFeature(DataType.FLOAT);
 
 					return continuousFeature;
 				}

diff --git a/pmml-sklearn/src/test/java/org/jpmml/sklearn/testing/ClassifierTest.java b/pmml-sklearn/src/test/java/org/jpmml/sklearn/testing/ClassifierTest.java
@@ -52,12 +52,22 @@ public List<Map<String, Object>> getOptionsMatrix(){
 				String algorithm = getAlgorithm();
 				String dataset = getDataset();
 
+				if((AUDIT).equals(dataset) || (IRIS).equals(dataset)){
+
+					if((DECISION_TREE).equals(algorithm) || (RANDOM_FOREST).equals(algorithm)){
+						Map<String, Object> options = new LinkedHashMap<>();
+						options.put(HasTreeOptions.OPTION_INPUT_FLOAT, new Boolean[]{false, true});
+
+						return OptionsUtil.generateOptionsMatrix(options);
+					}
+				} else
+
 				if((AUDIT_NA).equals(dataset) || (IRIS_NA).equals(dataset)){
 
 					if((RANDOM_FOREST).equals(algorithm)){
 						Map<String, Object> options = new LinkedHashMap<>();
 						options.put(HasTreeOptions.OPTION_ALLOW_MISSING, Boolean.TRUE);
-						options.put(HasTreeOptions.OPTION_COMPACT, new Boolean[]{true, false});
+						options.put(HasTreeOptions.OPTION_COMPACT, new Boolean[]{false, true});
 
 						return OptionsUtil.generateOptionsMatrix(options);
 					}

diff --git a/pmml-sklearn/src/test/java/org/jpmml/sklearn/testing/OutlierDetectorTest.java b/pmml-sklearn/src/test/java/org/jpmml/sklearn/testing/OutlierDetectorTest.java
@@ -52,6 +52,7 @@ public List<Map<String, Object>> getOptionsMatrix(){
 
 				if((ISOLATION_FOREST).equals(algorithm)){
 					Map<String, Object> options = new LinkedHashMap<>();
+					options.put(HasTreeOptions.OPTION_INPUT_FLOAT, new Boolean[]{false, true});
 					options.put(HasTreeOptions.OPTION_PRUNE, new Boolean[]{false, true});
 
 					return OptionsUtil.generateOptionsMatrix(options);

diff --git a/pmml-sklearn/src/test/java/org/jpmml/sklearn/testing/RegressorTest.java b/pmml-sklearn/src/test/java/org/jpmml/sklearn/testing/RegressorTest.java
@@ -50,12 +50,22 @@ public List<Map<String, Object>> getOptionsMatrix(){
 				String algorithm = getAlgorithm();
 				String dataset = getDataset();
 
+				if((AUTO).equals(dataset)){
+
+					if((DECISION_TREE).equals(algorithm) || (RANDOM_FOREST).equals(algorithm)){
+						Map<String, Object> options = new LinkedHashMap<>();
+						options.put(HasTreeOptions.OPTION_INPUT_FLOAT, new Boolean[]{false, true});
+
+						return OptionsUtil.generateOptionsMatrix(options);
+					}
+				} else
+
 				if((AUTO_NA).equals(dataset)){
 
 					if((RANDOM_FOREST).equals(algorithm)){
 						Map<String, Object> options = new LinkedHashMap<>();
 						options.put(HasTreeOptions.OPTION_ALLOW_MISSING, Boolean.TRUE);
-						options.put(HasTreeOptions.OPTION_COMPACT, new Boolean[]{true, false});
+						options.put(HasTreeOptions.OPTION_COMPACT, new Boolean[]{false, true});
 
 						return OptionsUtil.generateOptionsMatrix(options);
 					}