Java TreeModel类代码示例

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
@Override
public VisitorAction visit(TreeModel treeModel){
	TreeModel.MissingValueStrategy missingValueStrategy = treeModel.getMissingValueStrategy();
	TreeModel.NoTrueChildStrategy noTrueChildStrategy = treeModel.getNoTrueChildStrategy();
	TreeModel.SplitCharacteristic splitCharacteristic = treeModel.getSplitCharacteristic();

	if(!(TreeModel.MissingValueStrategy.DEFAULT_CHILD).equals(missingValueStrategy) || !(TreeModel.NoTrueChildStrategy.RETURN_NULL_PREDICTION).equals(noTrueChildStrategy) || !(TreeModel.SplitCharacteristic.BINARY_SPLIT).equals(splitCharacteristic)){
		throw new IllegalArgumentException();
	}

	treeModel
		.setMissingValueStrategy(TreeModel.MissingValueStrategy.NONE)
		.setNoTrueChildStrategy(TreeModel.NoTrueChildStrategy.RETURN_LAST_PREDICTION)
		.setSplitCharacteristic(TreeModel.SplitCharacteristic.MULTI_SPLIT);

	return super.visit(treeModel);
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
@Override
public MiningModel encodeModel(Schema schema){
	GBTClassificationModel model = getTransformer();

	String lossType = model.getLossType();
	switch(lossType){
		case "logistic":
			break;
		default:
			throw new IllegalArgumentException("Loss function " + lossType + " is not supported");
	}

	Schema segmentSchema = new Schema(new ContinuousLabel(null, DataType.DOUBLE), schema.getFeatures());

	List<TreeModel> treeModels = TreeModelUtil.encodeDecisionTreeEnsemble(model, segmentSchema);

	MiningModel miningModel = new MiningModel(MiningFunction.REGRESSION, ModelUtil.createMiningSchema(segmentSchema.getLabel()))
		.setSegmentation(MiningModelUtil.createSegmentation(Segmentation.MultipleModelMethod.WEIGHTED_SUM, treeModels, Doubles.asList(model.treeWeights())))
		.setOutput(ModelUtil.createPredictedOutput(FieldName.create("gbtValue"), OpType.CONTINUOUS, DataType.DOUBLE));

	return MiningModelUtil.createBinaryLogisticClassification(miningModel, 2d, 0d, RegressionModel.NormalizationMethod.LOGIT, false, schema);
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
static
public TreeModel encodeTreeModel(org.apache.spark.ml.tree.Node node, PredicateManager predicateManager, MiningFunction miningFunction, Schema schema){
	Node root = encodeNode(node, predicateManager, Collections.<FieldName, Set<String>>emptyMap(), miningFunction, schema)
		.setPredicate(new True());

	TreeModel treeModel = new TreeModel(miningFunction, ModelUtil.createMiningSchema(schema.getLabel()), root)
		.setSplitCharacteristic(TreeModel.SplitCharacteristic.BINARY_SPLIT);

	String compact = TreeModelOptions.COMPACT;
	if(compact != null && Boolean.valueOf(compact)){
		Visitor visitor = new TreeModelCompactor();

		visitor.applyTo(treeModel);
	}

	return treeModel;
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
static
public <E extends Estimator & HasTree> TreeModel encodeTreeModel(E estimator, PredicateManager predicateManager, MiningFunction miningFunction, Schema schema){
	Tree tree = estimator.getTree();

	int[] leftChildren = tree.getChildrenLeft();
	int[] rightChildren = tree.getChildrenRight();
	int[] features = tree.getFeature();
	double[] thresholds = tree.getThreshold();
	double[] values = tree.getValues();

	Node root = new Node()
		.setId("1")
		.setPredicate(new True());

	encodeNode(root, predicateManager, 0, leftChildren, rightChildren, features, thresholds, values, miningFunction, schema);

	TreeModel treeModel = new TreeModel(miningFunction, ModelUtil.createMiningSchema(schema.getLabel()), root)
		.setSplitCharacteristic(TreeModel.SplitCharacteristic.BINARY_SPLIT);

	return treeModel;
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
private MiningModel encodeRegression(RGenericVector ranger, Schema schema){
	RGenericVector forest = (RGenericVector)ranger.getValue("forest");

	ScoreEncoder scoreEncoder = new ScoreEncoder(){

		@Override
		public void encode(Node node, Number splitValue, RNumberVector<?> terminalClassCount){
			node.setScore(ValueUtil.formatValue(splitValue));
		}
	};

	List<TreeModel> treeModels = encodeForest(forest, MiningFunction.REGRESSION, scoreEncoder, schema);

	MiningModel miningModel = new MiningModel(MiningFunction.REGRESSION, ModelUtil.createMiningSchema(schema.getLabel()))
		.setSegmentation(MiningModelUtil.createSegmentation(Segmentation.MultipleModelMethod.AVERAGE, treeModels));

	return miningModel;
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
private List<TreeModel> encodeForest(RGenericVector forest, MiningFunction miningFunction, ScoreEncoder scoreEncoder, Schema schema){
	RNumberVector<?> numTrees = (RNumberVector<?>)forest.getValue("num.trees");
	RGenericVector childNodeIDs = (RGenericVector)forest.getValue("child.nodeIDs");
	RGenericVector splitVarIDs = (RGenericVector)forest.getValue("split.varIDs");
	RGenericVector splitValues = (RGenericVector)forest.getValue("split.values");
	RGenericVector terminalClassCounts = (RGenericVector)forest.getValue("terminal.class.counts", true);

	Schema segmentSchema = schema.toAnonymousSchema();

	List<TreeModel> treeModels = new ArrayList<>();

	for(int i = 0; i < ValueUtil.asInt(numTrees.asScalar()); i++){
		TreeModel treeModel = encodeTreeModel(miningFunction, scoreEncoder, (RGenericVector)childNodeIDs.getValue(i), (RNumberVector<?>)splitVarIDs.getValue(i), (RNumberVector<?>)splitValues.getValue(i), (terminalClassCounts != null ? (RGenericVector)terminalClassCounts.getValue(i) : null), segmentSchema);

		treeModels.add(treeModel);
	}

	return treeModels;
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
static
protected MiningModel createMiningModel(List<RegressionTree> regTrees, float base_score, Schema schema){
    ContinuousLabel continuousLabel = (ContinuousLabel)schema.getLabel();

    Schema segmentSchema = schema.toAnonymousSchema();

    List<TreeModel> treeModels = new ArrayList<>();

    for(RegressionTree regTree : regTrees){
        TreeModel treeModel = regTree.encodeTreeModel(segmentSchema);

        treeModels.add(treeModel);
    }

    MiningModel miningModel = new MiningModel(MiningFunction.REGRESSION, ModelUtil.createMiningSchema(continuousLabel))
            .setMathContext(MathContext.FLOAT)
            .setSegmentation(MiningModelUtil.createSegmentation(Segmentation.MultipleModelMethod.SUM, treeModels))
            .setTargets(ModelUtil.createRescaleTargets(null, ValueUtil.floatToDouble(base_score), continuousLabel));

    return miningModel;
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
private Trail handleNoTrueChild(Trail trail){
	TreeModel treeModel = getModel();

	TreeModel.NoTrueChildStrategy noTrueChildStrategy = treeModel.getNoTrueChildStrategy();
	switch(noTrueChildStrategy){
		case RETURN_NULL_PREDICTION:
			return trail.selectNull();
		case RETURN_LAST_PREDICTION:
			Node lastPrediction = trail.getLastPrediction();

			// "Return the parent Node only if it specifies a score attribute"
			if(lastPrediction.hasScore()){
				return trail.selectLastPrediction();
			}
			return trail.selectNull();
		default:
			throw new UnsupportedAttributeException(treeModel, noTrueChildStrategy);
	}
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
/**
 * @param parent The parent Node of the Node that evaluated to the missing value.
 * @param node The Node that evaluated to the missing value.
 */
private Trail handleMissingValue(Trail trail, Node parent, Node node, EvaluationContext context){
	TreeModel treeModel = getModel();

	TreeModel.MissingValueStrategy missingValueStrategy = treeModel.getMissingValueStrategy();
	switch(missingValueStrategy){
		case NULL_PREDICTION:
			return trail.selectNull();
		case LAST_PREDICTION:
			return trail.selectLastPrediction();
		case DEFAULT_CHILD:
			return handleDefaultChild(trail, parent, context);
		case NONE:
			return null;
		default:
			throw new UnsupportedAttributeException(treeModel, missingValueStrategy);
	}
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
@Test
public void defaultChildSinglePenalty() throws Exception {
	Map<FieldName, ?> arguments = createArguments("outlook", null, "temperature", 40d, "humidity", 70d);

	NodeScoreDistribution<?> targetValue = evaluate(TreeModel.MissingValueStrategy.DEFAULT_CHILD, 0.8d, arguments);

	assertEquals("4", targetValue.getEntityId());

	assertEquals((Double)0.4d, targetValue.getProbability("will play"));
	assertEquals((Double)0d, targetValue.getProbability("may play"));
	assertEquals((Double)0.6d, targetValue.getProbability("no play"));

	double missingValuePenatly = 0.8d;

	assertEquals((Double)(0.4d * missingValuePenatly), targetValue.getConfidence("will play"));
	assertEquals((Double)(0d * missingValuePenatly), targetValue.getConfidence("may play"));
	assertEquals((Double)(0.6d * missingValuePenatly), targetValue.getConfidence("no play"));
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
@Test
public void defaultChildMultiplePenalties() throws Exception {
	Map<FieldName, ?> arguments = createArguments("outlook", null, "temperature", null, "humidity", 70d);

	NodeScoreDistribution<?> targetValue = evaluate(TreeModel.MissingValueStrategy.DEFAULT_CHILD, 0.8d, arguments);

	assertEquals("3", targetValue.getEntityId());

	assertEquals((Double)0.9d, targetValue.getProbability("will play"));
	assertEquals((Double)0.05d, targetValue.getProbability("may play"));
	assertEquals((Double)0.05d, targetValue.getProbability("no play"));

	double missingValuePenalty = (0.8d * 0.8d);

	assertEquals((Double)(0.9d * missingValuePenalty), targetValue.getConfidence("will play"));
	assertEquals((Double)(0.05d * missingValuePenalty), targetValue.getConfidence("may play"));
	assertEquals((Double)(0.05d * missingValuePenalty), targetValue.getConfidence("no play"));
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
static
private TreeModelEvaluator createTreeModelEvaluator(MiningFunction miningFunction, MathContext mathContext, Target target){
	Node root = new Node()
		.setPredicate(new False());

	Targets targets = new Targets()
		.addTargets(target);

	TreeModel treeModel = new TreeModel(miningFunction, new MiningSchema(), root)
		.setSplitCharacteristic(TreeModel.SplitCharacteristic.BINARY_SPLIT)
		.setMathContext(mathContext)
		.setTargets(targets);

	PMML pmml = new PMML(Version.PMML_4_3.getVersion(), new Header(), new DataDictionary())
		.addModels(treeModel);

	return new TreeModelEvaluator(pmml);
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
@Override
public VisitorAction visit(Node node){
	TreeModel treeModel = getTreeModel();

	MiningFunction miningFunction = treeModel.getMiningFunction();
	switch(miningFunction){
		case REGRESSION:
			if(node.hasScoreDistributions()){
				List<ScoreDistribution> scoreDistributions = node.getScoreDistributions();

				scoreDistributions.clear();
			}
			break;
		default:
			break;
	}

	return super.visit(node);
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
@Test
public void clean(){
	Node node = new Node()
		.setPredicate(new True())
		.setScore("1")
		.addScoreDistributions(new ScoreDistribution("0", 0), new ScoreDistribution("1", 100));

	TreeModel treeModel = new TreeModel(MiningFunction.CLASSIFICATION, new MiningSchema(), node);

	ScoreDistributionCleaner cleaner = new ScoreDistributionCleaner();
	cleaner.applyTo(treeModel);

	assertTrue(node.hasScoreDistributions());

	treeModel.setMiningFunction(MiningFunction.REGRESSION);

	cleaner.applyTo(treeModel);

	assertFalse(node.hasScoreDistributions());
}
 

import org.dmg.pmml.tree.TreeModel; //导入依赖的package包/类
@Test
public void intern(){
	ScoreDistribution left = new ScoreDistribution("event", 0.33d);
	ScoreDistribution right = new ScoreDistribution("event", 0.33d);

	Node leftChild = createNode(left);
	Node rightChild = createNode(right);

	Node root = new Node()
		.setPredicate(new True())
		.addNodes(leftChild, rightChild);

	TreeModel treeModel = new TreeModel()
		.setNode(root);

	for(int i = 0; i < 2; i++){
		assertNotSame((leftChild.getScoreDistributions()).get(i), (rightChild.getScoreDistributions()).get(i));
	}

	intern(treeModel);

	for(int i = 0; i < 2; i++){
		assertSame((leftChild.getScoreDistributions()).get(i), (rightChild.getScoreDistributions()).get(i));
	}
}