org.dmg.pmml.mining.MiningModel Java Examples

@Override
public void pushParent(PMMLObject parent){
	super.pushParent(parent);

	if(parent instanceof MiningModel){
		MiningModel miningModel = (MiningModel)parent;

		processMiningModel(miningModel);
	} else

	if(parent instanceof Model){
		Model model = (Model)parent;

		processModel(model);
	}
}
 

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;
}
 

static
public MiningModel createModelChain(List<? extends Model> models, Schema schema){

    if(models.size() < 1){
        throw new IllegalArgumentException();
    }

    Segmentation segmentation = createSegmentation(Segmentation.MultipleModelMethod.MODEL_CHAIN, models);

    Model lastModel = Iterables.getLast(models);

    MiningModel miningModel = new MiningModel(lastModel.getMiningFunction(), ModelUtil.createMiningSchema(schema.getLabel()))
            .setMathContext(ModelUtil.simplifyMathContext(lastModel.getMathContext()))
            .setSegmentation(segmentation);

    return miningModel;
}
 

@Override
public MiningModel encodeMiningModel(List<Tree> trees, Integer numIteration, Schema schema){
	Schema segmentSchema = schema.toAnonymousRegressorSchema(DataType.DOUBLE);

	List<MiningModel> miningModels = new ArrayList<>();

	CategoricalLabel categoricalLabel = (CategoricalLabel)schema.getLabel();

	for(int i = 0, rows = categoricalLabel.size(), columns = (trees.size() / rows); i < rows; i++){
		MiningModel miningModel = createMiningModel(FortranMatrixUtil.getRow(trees, rows, columns, i), numIteration, segmentSchema)
			.setOutput(ModelUtil.createPredictedOutput(FieldName.create("lgbmValue(" + categoricalLabel.getValue(i) + ")"), OpType.CONTINUOUS, DataType.DOUBLE));

		miningModels.add(miningModel);
	}

	return MiningModelUtil.createClassification(miningModels, RegressionModel.NormalizationMethod.SOFTMAX, true, schema);
}
 

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;
}
 

@Override
public MiningModel encodeModel(Schema schema){
	RGenericVector randomForest = getObject();

	RStringVector type = randomForest.getStringElement("type");
	RGenericVector forest = randomForest.getGenericElement("forest");

	switch(type.asScalar()){
		case "regression":
			return encodeRegression(forest, schema);
		case "classification":
			return encodeClassification(forest, schema);
		default:
			throw new IllegalArgumentException();
	}
}
 

private void processMiningModel(MiningModel miningModel){
	Segmentation segmentation = miningModel.getSegmentation();

	if(segmentation != null){
		Segmentation.MultipleModelMethod multipleModelMethod = segmentation.getMultipleModelMethod();

		switch(multipleModelMethod){
			case SELECT_FIRST:
			case SELECT_ALL:
			case MODEL_CHAIN:
				{
					this.targetDataTypes.push(Collections.singletonMap(Evaluator.DEFAULT_TARGET_NAME, null));

					this.dataType = null;

					return;
				}
			default:
				break;
		}
	}

	processModel(miningModel);
}
 

@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 = schema.toAnonymousRegressorSchema(DataType.DOUBLE);

	List<TreeModel> treeModels = TreeModelUtil.encodeDecisionTreeEnsemble(this, 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);
}
 

private MiningModel encodeRegression(RGenericVector ranger, Schema schema){
	RGenericVector forest = ranger.getGenericElement("forest");

	ScoreEncoder scoreEncoder = new ScoreEncoder(){

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

			return node;
		}
	};

	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;
}
 

@Override
public MiningModel encodeModel(Schema schema){
	RGenericVector ranger = getObject();

	RStringVector treetype = ranger.getStringElement("treetype");

	switch(treetype.asScalar()){
		case "Regression":
			return encodeRegression(ranger, schema);
		case "Classification":
			return encodeClassification(ranger, schema);
		case "Probability estimation":
			return encodeProbabilityForest(ranger, schema);
		default:
			throw new IllegalArgumentException();
	}
}
 

@Override
public Model encodeModel(Schema schema){
	Classifier gbdt = getGBDT();
	MultiOneHotEncoder ohe = getOHE();
	LinearClassifier lr = getLR();

	CategoricalLabel categoricalLabel = (CategoricalLabel)schema.getLabel();

	SchemaUtil.checkSize(2, categoricalLabel);

	List<? extends Number> coef = lr.getCoef();
	List<? extends Number> intercept = lr.getIntercept();

	Schema segmentSchema = schema.toAnonymousSchema();

	MiningModel miningModel = GBDTUtil.encodeModel(gbdt, ohe, coef, Iterables.getOnlyElement(intercept), segmentSchema)
		.setOutput(ModelUtil.createPredictedOutput(FieldName.create("decisionFunction"), OpType.CONTINUOUS, DataType.DOUBLE));

	return MiningModelUtil.createBinaryLogisticClassification(miningModel, 1d, 0d, RegressionModel.NormalizationMethod.LOGIT, lr.hasProbabilityDistribution(), schema);
}
 

@Override
public MiningModel encodeMiningModel(List<RegTree> trees, List<Float> weights, float base_score, Integer ntreeLimit, Schema schema){
	Schema segmentSchema = schema.toAnonymousRegressorSchema(DataType.FLOAT);

	List<MiningModel> miningModels = new ArrayList<>();

	CategoricalLabel categoricalLabel = (CategoricalLabel)schema.getLabel();

	for(int i = 0, columns = categoricalLabel.size(), rows = (trees.size() / columns); i < columns; i++){
		MiningModel miningModel = createMiningModel(CMatrixUtil.getColumn(trees, rows, columns, i), (weights != null) ? CMatrixUtil.getColumn(weights, rows, columns, i) : null, base_score, ntreeLimit, segmentSchema)
			.setOutput(ModelUtil.createPredictedOutput(FieldName.create("xgbValue(" + categoricalLabel.getValue(i) + ")"), OpType.CONTINUOUS, DataType.FLOAT));

		miningModels.add(miningModel);
	}

	return MiningModelUtil.createClassification(miningModels, RegressionModel.NormalizationMethod.SOFTMAX, true, schema);
}
 

private List<OutputField> createNestedOutputFields(){
	MiningModel miningModel = getModel();

	Segmentation segmentation = miningModel.getSegmentation();

	List<Segment> segments = segmentation.getSegments();

	Segmentation.MultipleModelMethod multipleModelMethod = segmentation.getMultipleModelMethod();
	switch(multipleModelMethod){
		case SELECT_FIRST:
			return createNestedOutputFields(getActiveHead(segments));
		case SELECT_ALL:
			// Ignored
			break;
		case MODEL_CHAIN:
			return createNestedOutputFields(getActiveTail(segments));
		default:
			break;
	}

	return Collections.emptyList();
}
 

@Override
public MiningModel encodeModel(Schema schema){
	List<? extends Regressor> estimators = getEstimators();
	List<? extends Number> estimatorWeights = getEstimatorWeights();

	Schema segmentSchema = schema.toAnonymousSchema();

	List<Model> models = new ArrayList<>();

	for(Regressor estimator : estimators){
		Model model = estimator.encodeModel(segmentSchema);

		models.add(model);
	}

	MiningModel miningModel = new MiningModel(MiningFunction.REGRESSION, ModelUtil.createMiningSchema(schema.getLabel()))
		.setSegmentation(MiningModelUtil.createSegmentation(MultipleModelMethod.WEIGHTED_MEDIAN, models, estimatorWeights));

	return miningModel;
}
 

@Override
public MiningModel encodeModel(Schema schema){
	List<? extends Classifier> estimators = getEstimators();
	List<List<Integer>> estimatorsFeatures = getEstimatorsFeatures();

	Segmentation.MultipleModelMethod multipleModelMethod = Segmentation.MultipleModelMethod.AVERAGE;

	for(Classifier estimator : estimators){

		if(!estimator.hasProbabilityDistribution()){
			multipleModelMethod = Segmentation.MultipleModelMethod.MAJORITY_VOTE;

			break;
		}
	}

	MiningModel miningModel = BaggingUtil.encodeBagging(estimators, estimatorsFeatures, multipleModelMethod, MiningFunction.CLASSIFICATION, schema)
		.setOutput(ModelUtil.createProbabilityOutput(DataType.DOUBLE, (CategoricalLabel)schema.getLabel()));

	return miningModel;
}
 

@Override
public DataField getDefaultDataField(){
	MiningModel miningModel = getModel();

	Segmentation segmentation = miningModel.getSegmentation();

	Segmentation.MultipleModelMethod multipleModelMethod = segmentation.getMultipleModelMethod();
	switch(multipleModelMethod){
		case SELECT_FIRST:
		case SELECT_ALL:
		case MODEL_CHAIN:
			return null;
		default:
			return super.getDefaultDataField();
	}
}
 

@Override
public MiningModel encodeMiningModel(List<RegTree> trees, List<Float> weights, float base_score, Integer ntreeLimit, Schema schema){
	Schema segmentSchema = schema.toAnonymousRegressorSchema(DataType.FLOAT);

	Transformation transformation = new FunctionTransformation(PMMLFunctions.THRESHOLD){

		@Override
		public FieldName getName(FieldName name){
			return FieldName.create("hinge(" + name + ")");
		}

		@Override
		public Expression createExpression(FieldRef fieldRef){
			Apply apply = (Apply)super.createExpression(fieldRef);

			apply.addExpressions(PMMLUtil.createConstant(0f));

			return apply;
		}
	};

	MiningModel miningModel = createMiningModel(trees, weights, base_score, ntreeLimit, segmentSchema)
		.setOutput(ModelUtil.createPredictedOutput(FieldName.create("xgbValue"), OpType.CONTINUOUS, DataType.FLOAT, transformation));

	return MiningModelUtil.createBinaryLogisticClassification(miningModel, 1d, 0d, RegressionModel.NormalizationMethod.NONE, true, schema);
}
 

@Override
public Model encodeModel(Schema schema){
	List<? extends Regressor> estimators = getEstimators();
	List<? extends Number> weights = getWeights();

	List<Model> models = new ArrayList<>();

	for(Regressor estimator : estimators){
		Model model = estimator.encodeModel(schema);

		models.add(model);
	}

	Segmentation.MultipleModelMethod multipleModelMethod = (weights != null && weights.size() > 0 ? Segmentation.MultipleModelMethod.WEIGHTED_AVERAGE : Segmentation.MultipleModelMethod.AVERAGE);

	MiningModel miningModel = new MiningModel(MiningFunction.REGRESSION, ModelUtil.createMiningSchema(schema.getLabel()))
		.setSegmentation(MiningModelUtil.createSegmentation(multipleModelMethod, models, weights));

	return miningModel;
}
 

@Override
public Model encodeModel(Schema schema){
	RGenericVector ada = getObject();

	RGenericVector model = ada.getGenericElement("model");

	RGenericVector trees = model.getGenericElement("trees");
	RDoubleVector alpha = model.getDoubleElement("alpha");

	List<TreeModel> treeModels = encodeTreeModels(trees);

	MiningModel miningModel = new MiningModel(MiningFunction.REGRESSION, ModelUtil.createMiningSchema(null))
		.setSegmentation(MiningModelUtil.createSegmentation(Segmentation.MultipleModelMethod.WEIGHTED_SUM, treeModels, alpha.getValues()))
		.setOutput(ModelUtil.createPredictedOutput(FieldName.create("adaValue"), OpType.CONTINUOUS, DataType.DOUBLE));

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

@Override
public MiningModel encodeModel(Schema schema){
	HasDefaultValue init = getInit();
	Number learningRate = getLearningRate();

	return GradientBoostingUtil.encodeGradientBoosting(this, init.getDefaultValue(), learningRate, schema);
}
 

@Override
public MiningModel encodeModel(Schema schema){
	List<List<TreePredictor>> predictors = getPredictors();
	Number baselinePrediction = getBaselinePrediction();

	return HistGradientBoostingUtil.encodeHistGradientBoosting(predictors, Collections.singletonList(baselinePrediction), 0, schema);
}
 

static
public MiningModel encodeHistGradientBoosting(List<List<TreePredictor>> predictors, List<? extends Number> baselinePredictions, int column, Schema schema){
	List<TreePredictor> treePredictors = predictors.stream()
		.map(predictor -> predictor.get(column))
		.collect(Collectors.toList());

	Number baselinePrediction = baselinePredictions.get(column);

	return encodeHistGradientBoosting(treePredictors, baselinePrediction, schema);
}
 

private MiningModel encodeBinaryClassification(List<TreeModel> treeModels, Double initF, double coefficient, Schema schema){
	Schema segmentSchema = schema.toAnonymousRegressorSchema(DataType.DOUBLE);

	MiningModel miningModel = createMiningModel(treeModels, initF, segmentSchema)
		.setOutput(ModelUtil.createPredictedOutput(FieldName.create("gbmValue"), OpType.CONTINUOUS, DataType.DOUBLE));

	return MiningModelUtil.createBinaryLogisticClassification(miningModel, -coefficient, 0d, RegressionModel.NormalizationMethod.LOGIT, true, schema);
}
 

private MiningModel encodeMiningModel(RStringVector distributionName, List<TreeModel> treeModels, Double initF, Schema schema){

		switch(distributionName.asScalar()){
			case "gaussian":
				return encodeRegression(treeModels, initF, schema);
			case "adaboost":
				return encodeBinaryClassification(treeModels, initF, -2d, schema);
			case "bernoulli":
				return encodeBinaryClassification(treeModels, initF, -1d, schema);
			case "multinomial":
				return encodeMultinomialClassification(treeModels, initF, schema);
			default:
				throw new IllegalArgumentException();
		}
	}
 

@Override
public MiningModel encodeModel(Schema schema){
	XGBoostRegressionModel model = getTransformer();

	Booster booster = model.nativeBooster();

	return BoosterUtil.encodeBooster(this, booster, schema);
}
 

static
private MiningModel encodeModel(MiningFunction miningFunction, List<Object[]> steps, Schema schema){

	if(steps.size() < 1){
		throw new IllegalArgumentException();
	}

	Label label = schema.getLabel();
	List<? extends Feature> features = schema.getFeatures();

	Segmentation segmentation = new Segmentation(Segmentation.MultipleModelMethod.SELECT_FIRST, null);

	Scope scope = new DataFrameScope(FieldName.create("X"), features);

	for(Object[] step : steps){
		String name = TupleUtil.extractElement(step, 0, String.class);
		Estimator estimator = TupleUtil.extractElement(step, 1, Estimator.class);
		String predicate = TupleUtil.extractElement(step, 2, String.class);

		if(!(miningFunction).equals(estimator.getMiningFunction())){
			throw new IllegalArgumentException();
		}

		Predicate pmmlPredicate = PredicateTranslator.translate(predicate, scope);

		Model model = estimator.encodeModel(schema);

		Segment segment = new Segment(pmmlPredicate, model)
			.setId(name);

		segmentation.addSegments(segment);
	}

	MiningModel miningModel = new MiningModel(miningFunction, ModelUtil.createMiningSchema(label))
		.setSegmentation(segmentation);

	return miningModel;
}
 

static
public <E extends Estimator & HasEstimatorEnsemble<TreeRegressor> & HasTreeOptions> MiningModel encodeGradientBoosting(E estimator, Number initialPrediction, Number learningRate, Schema schema){
	ContinuousLabel continuousLabel = (ContinuousLabel)schema.getLabel();

	List<TreeModel> treeModels = TreeUtil.encodeTreeModelEnsemble(estimator, MiningFunction.REGRESSION, schema);

	MiningModel miningModel = new MiningModel(MiningFunction.REGRESSION, ModelUtil.createMiningSchema(continuousLabel))
		.setSegmentation(MiningModelUtil.createSegmentation(Segmentation.MultipleModelMethod.SUM, treeModels))
		.setTargets(ModelUtil.createRescaleTargets(learningRate, initialPrediction, continuousLabel));

	return TreeUtil.transform(estimator, miningModel);
}
 

public static PMML encodePMML(FieldName targetField, List<String> targetCategories, FeatureList featureList, List<List<RegressionTree>> regTrees, float base_score, int numClasses){
    LKBoostEncoder encoder = new LKBoostEncoder();

    if(targetField == null){
        targetField = FieldName.create("_target");
    }

    Label label = encodeLabel(targetField, targetCategories, encoder, numClasses);

    //todo
    List<Feature> features = new ArrayList<>();
    for (int i=0;i<featureList.size();i++){
        FieldName fieldName = new FieldName("feature_"+i);
        DataField dataField = encoder.createDataField(fieldName, OpType.CONTINUOUS, DataType.FLOAT);
        Feature feature = new ContinuousFeature(encoder, dataField);
        features.add(feature);
    }


    Schema schema = new Schema(label, features);

    MiningModel miningModel = encodeMiningModel(regTrees, base_score, schema);

    PMML pmml = encoder.encodePMML(miningModel);

    return pmml;
}
 

static
public <E extends Estimator & HasEstimatorEnsemble<T> & HasTreeOptions, T extends Estimator & HasTree> MiningModel encodeBaseForest(E estimator, Segmentation.MultipleModelMethod multipleModelMethod, MiningFunction miningFunction, Schema schema){
	List<TreeModel> treeModels = TreeUtil.encodeTreeModelEnsemble(estimator, miningFunction, schema);

	MiningModel miningModel = new MiningModel(miningFunction, ModelUtil.createMiningSchema(schema.getLabel()))
		.setSegmentation(MiningModelUtil.createSegmentation(multipleModelMethod, treeModels));

	return TreeUtil.transform(estimator, miningModel);
}
 

public static PMML encodePMML(FieldName targetField, List<String> targetCategories, FeatureList featureList, List<RegressionTree> regTrees, float base_score){
    LSBoostEncoder encoder = new LSBoostEncoder();

    if(targetField == null){
        targetField = FieldName.create("_target");
    }

    Label label = encodeLabel(targetField, targetCategories, encoder);

    //todo
    List<Feature> features = new ArrayList<>();
    for (int i=0;i<featureList.size();i++){
        FieldName fieldName = new FieldName("feature_"+i);
        DataField dataField = encoder.createDataField(fieldName, OpType.CONTINUOUS, DataType.FLOAT);
        Feature feature = new ContinuousFeature(encoder, dataField);
        features.add(feature);
    }


    Schema schema = new Schema(label, features);

    MiningModel miningModel = encodeMiningModel(regTrees, base_score, schema);

    PMML pmml = encoder.encodePMML(miningModel);

    return pmml;
}