Java PredictionModel类代码示例

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
/**
 * Applies the applier and evaluator (= second subprocess). In order to reuse possibly created
 * predicted label attributes, we do the following: We compare the predicted label of
 * <code>testSet</code> before and after applying the inner operator. If it changed, the
 * predicted label is removed again. No outer operator could ever see it. The same applies for
 * the confidence attributes in case of classification learning.
 */
protected final void evaluate(ExampleSet testSet) throws OperatorException {
	Attribute predictedBefore = testSet.getAttributes().getPredictedLabel();

	applyProcessExampleSetOutput.deliver(testSet);
	applyProcessModelOutput.deliver(trainingProcessModelInput.getData(IOObject.class));
	throughExtender.passDataThrough();

	executeEvaluator();

	Tools.buildAverages(applyProcessPerformancePortExtender);

	Attribute predictedAfter = testSet.getAttributes().getPredictedLabel();
	// remove predicted label and confidence attributes if there is a new prediction which is
	// not equal to an old one
	if (predictedAfter != null
			&& (predictedBefore == null || predictedBefore.getTableIndex() != predictedAfter.getTableIndex())) {
		PredictionModel.removePredictedLabel(testSet);
	}
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
/** Constructor. */
public BayesianBoosting(OperatorDescription description) {
	super(description);

	modelInput.addPrecondition(new SimplePrecondition(modelInput, new PredictionModelMetaData(PredictionModel.class,
			new ExampleSetMetaData()), false));

	addValue(new ValueDouble("performance", "The performance.") {

		@Override
		public double getDoubleValue() {
			return performance;
		}
	});
	addValue(new ValueDouble("iteration", "The current iteration.") {

		@Override
		public double getDoubleValue() {
			return currentIteration;
		}
	});
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
/**
 * This methods replaces the labels of the given example set with the label residuals after
 * using the given model. Please note that the label column will be overwritten and the original
 * label should be stored!
 */
private void residualReplace(ExampleSet exampleSet, Model model, boolean shrinkage) throws OperatorException {
	ExampleSet resultSet = model.apply(exampleSet);
	Attribute label = exampleSet.getAttributes().getLabel();
	Iterator<Example> originalReader = exampleSet.iterator();
	Iterator<Example> predictionReader = resultSet.iterator();
	while ((originalReader.hasNext()) && (predictionReader.hasNext())) {
		Example originalExample = originalReader.next();
		Example predictionExample = predictionReader.next();
		double prediction = predictionExample.getPredictedLabel();
		if (shrinkage) {
			prediction *= getParameterAsDouble(PARAMETER_SHRINKAGE);
		}
		double residual = originalExample.getLabel() - prediction;
		originalExample.setValue(label, residual);
	}
	PredictionModel.removePredictedLabel(resultSet);
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
/**
 * This methods replaces the labels of the given example set with the label residuals after
 * using the given model. Please note that the label column will be overwritten and the original
 * label should be stored!
 */
private void residualReplace(ExampleSet exampleSet, Model model, boolean shrinkage) throws OperatorException {
	ExampleSet resultSet = model.apply(exampleSet);
	Attribute label = exampleSet.getAttributes().getLabel();
	Iterator<Example> originalReader = exampleSet.iterator();
	Iterator<Example> predictionReader = resultSet.iterator();
	double shrinkageValue = 0;
	if (shrinkage) {
		shrinkageValue = getParameterAsDouble(PARAMETER_SHRINKAGE);
	}
	while (originalReader.hasNext() && predictionReader.hasNext()) {
		Example originalExample = originalReader.next();
		Example predictionExample = predictionReader.next();
		double prediction = predictionExample.getPredictedLabel();
		if (shrinkage) {
			prediction *= shrinkageValue;
		}
		double residual = originalExample.getLabel() - prediction;
		originalExample.setValue(label, residual);
	}
	PredictionModel.removePredictedLabel(resultSet);
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
/**
 * Applies the applier and evaluator (= second subprocess).
 * In order to reuse possibly created predicted label attributes, we do the
 * following: We compare the predicted label of <code>testSet</code>
 * before and after applying the inner operator. If it changed, the
 * predicted label is removed again. No outer operator could ever see it.
 * The same applies for the confidence attributes in case of classification
 * learning.
 */
protected final void evaluate(ExampleSet testSet) throws OperatorException {
    Attribute predictedBefore = testSet.getAttributes().getPredictedLabel();

    applyProcessExampleSetOutput.deliver(testSet);
    applyProcessModelOutput.deliver(trainingProcessModelInput.getData(IOObject.class));
    throughExtender.passDataThrough();

    executeEvaluator();

    Tools.buildAverages(applyProcessPerformancePortExtender);

    Attribute predictedAfter = testSet.getAttributes().getPredictedLabel();
    // remove predicted label and confidence attributes if there is a new prediction which is not equal to an old one
    if ((predictedAfter != null) && ((predictedBefore == null) ||
            (predictedBefore.getTableIndex() != predictedAfter.getTableIndex()))) {
        PredictionModel.removePredictedLabel(testSet);
    }
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
/** Constructor. */
public BayesianBoosting(OperatorDescription description) {
	super(description);

	modelInput.addPrecondition(new SimplePrecondition(modelInput, new PredictionModelMetaData(PredictionModel.class, new ExampleSetMetaData()), false));

	addValue(new ValueDouble("performance", "The performance.") {
		@Override
		public double getDoubleValue() {
			return performance;
		}
	});
	addValue(new ValueDouble("iteration", "The current iteration.") {
		@Override
		public double getDoubleValue() {
			return currentIteration;
		}
	});
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
/**
 * This methods replaces the labels of the given example set with the label residuals after using the given model.
 * Please note that the label column will be overwritten and the original label should be stored!
 */
private void residualReplace(ExampleSet exampleSet, Model model, boolean shrinkage) throws OperatorException {
	ExampleSet resultSet = model.apply(exampleSet);
	Attribute label = exampleSet.getAttributes().getLabel();
	Iterator<Example> originalReader = exampleSet.iterator();
	Iterator<Example> predictionReader = resultSet.iterator();
	while ((originalReader.hasNext()) && (predictionReader.hasNext())) {
		Example originalExample = originalReader.next();
		Example predictionExample = predictionReader.next();
		double prediction = predictionExample.getPredictedLabel();
		if (shrinkage)
			prediction *= getParameterAsDouble(PARAMETER_SHRINKAGE);
		double residual = originalExample.getLabel() - prediction;
		originalExample.setValue(label, residual);
	}
	PredictionModel.removePredictedLabel(resultSet);
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
@Override
public void doWork() throws OperatorException {
	ExampleSet exampleSet = exampleSetInput.getData(ExampleSet.class);
	Model model = modelInput.getData(Model.class);

	// some checks
	if (exampleSet.getAttributes().getLabel() == null) {
		throw new UserError(this, 105, new Object[0]);
	}
	if (exampleSet.getAttributes().size() == 0) {
		throw new UserError(this, 106, new Object[0]);
	}

	final Attribute label = this.extractLabel(model, exampleSet);

	PlattParameters plattParams;
	{
		ExampleSet calibrationSet = (ExampleSet) exampleSet.clone();
		calibrationSet = model.apply(calibrationSet);
		plattParams = computeParameters(calibrationSet, label);
		PredictionModel.removePredictedLabel(calibrationSet);
	}

	PlattScalingModel scalingModel = new PlattScalingModel(exampleSet, model, plattParams);

	exampleSetOutput.deliver(scalingModel.apply(exampleSet));
	modelOutput.deliver(scalingModel);
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
private Attribute extractLabel(Model model, ExampleSet exampleSet) {
	if (model instanceof PredictionModel) {
		return ((PredictionModel) model).getLabel();
	}
	logWarning("Could not find label in model for Platt's Scaling, using Label of provided ExampleSet instead.");
	return exampleSet.getAttributes().getLabel();
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
@Override
public void doWork() throws OperatorException {
	ExampleSet exampleSet = exampleSetInput.getData(ExampleSet.class);
	if (exampleSet.getAttributes().getLabel() == null) {
		throw new UserError(this, 105);
	}
	if (!exampleSet.getAttributes().getLabel().isNominal()) {
		throw new UserError(this, 101, "ROC Charts", exampleSet.getAttributes().getLabel());
	}
	if (exampleSet.getAttributes().getLabel().getMapping().getValues().size() != 2) {
		throw new UserError(this, 114, "ROC Charts", exampleSet.getAttributes().getLabel());
	}

	if (exampleSet.getAttributes().getPredictedLabel() != null && getParameterAsBoolean(PARAMETER_USE_MODEL)) {
		getLogger().warning("Input example already has a predicted label which will be removed.");
		PredictionModel.removePredictedLabel(exampleSet);
	}
	if (exampleSet.getAttributes().getPredictedLabel() == null && !getParameterAsBoolean(PARAMETER_USE_MODEL)) {
		throw new UserError(this, 107);
	}
	Model model = null;
	if (getParameterAsBoolean(PARAMETER_USE_MODEL)) {
		model = modelInput.getData(Model.class);
		exampleSet = model.apply(exampleSet);
	}
	if (exampleSet.getAttributes().getPredictedLabel() == null) {
		throw new UserError(this, 107);
	}

	ROCDataGenerator rocDataGenerator = new ROCDataGenerator(1.0d, 1.0d);
	ROCData rocPoints = rocDataGenerator.createROCData(exampleSet, getParameterAsBoolean(PARAMETER_USE_EXAMPLE_WEIGHTS),
			ROCBias.getROCBiasParameter(this));
	rocDataGenerator.createROCPlotDialog(rocPoints);

	PredictionModel.removePredictedLabel(exampleSet);

	modelOutput.deliver(model);
	exampleSetOutput.deliver(exampleSet);
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
@Override
public void doWork() throws OperatorException {
	ExampleSet exampleSet = exampleSetInput.getData(ExampleSet.class);
	Model model = modelInput.getData(Model.class);

	if (exampleSet.getAttributes().getLabel() == null) {
		throw new UserError(this, 105);
	}
	if (!exampleSet.getAttributes().getLabel().isNominal()) {
		throw new UserError(this, 101, "Lift Charts", exampleSet.getAttributes().getLabel());
	}
	if (exampleSet.getAttributes().getLabel().getMapping().getValues().size() != 2) {
		throw new UserError(this, 114, "Lift Charts", exampleSet.getAttributes().getLabel());
	}

	ExampleSet workingSet = (ExampleSet) exampleSet.clone();
	if (workingSet.getAttributes().getPredictedLabel() != null) {
		PredictionModel.removePredictedLabel(workingSet);
	}

	workingSet = model.apply(workingSet);
	if (workingSet.getAttributes().getPredictedLabel() == null) {
		throw new UserError(this, 107);
	}

	LiftDataGenerator liftDataGenerator = new LiftDataGenerator();
	List<double[]> liftPoints = liftDataGenerator.createLiftDataList(workingSet);
	liftDataGenerator.createLiftChartPlot(liftPoints);

	PredictionModel.removePredictedLabel(workingSet);

	exampleSetOutput.deliver(exampleSet);
	modelOutput.deliver(model);
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
/**
 * Helper method applying the start model and adding it to the modelInfo collection
 */
private void applyPriorModel(ExampleSet trainingSet, List<BayBoostBaseModelInfo> modelInfo) throws OperatorException {
	// If the input contains a model already, initialise the example weights.
	if (this.startModel != null) {

		ExampleSet resultSet = this.startModel.apply((ExampleSet) trainingSet.clone());

		// Initial values and the input model are stored in the output model.
		WeightedPerformanceMeasures wp = new WeightedPerformanceMeasures(resultSet);

		this.reweightExamples(wp, resultSet);
		modelInfo.add(new BayBoostBaseModelInfo(this.startModel, wp.getContingencyMatrix()));
		PredictionModel.removePredictedLabel(resultSet);
	}
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
@Override
public ExampleSet performPrediction(ExampleSet exampleSet, Attribute predictedLabel) throws OperatorException {
	// apply default model
	exampleSet = defaultModel.apply(exampleSet);
	double[] predictions = new double[exampleSet.size()];
	Iterator<Example> e = exampleSet.iterator();
	int counter = 0;
	while (e.hasNext()) {
		predictions[counter++] = e.next().getPredictedLabel();
	}
	PredictionModel.removePredictedLabel(exampleSet);

	// apply all models to the example set sum up the predictions
	for (int i = 0; i < residualModels.length; i++) {
		exampleSet = residualModels[i].apply(exampleSet);
		e = exampleSet.iterator();
		counter = 0;
		while (e.hasNext()) {
			predictions[counter++] += shrinkage * e.next().getPredictedLabel();
		}
		PredictionModel.removePredictedLabel(exampleSet);
	}

	// set final predictions
	e = exampleSet.iterator();
	counter = 0;
	Attribute newPredictedLabel = createPredictedLabel(exampleSet, getLabel());
	while (e.hasNext()) {
		e.next().setValue(newPredictedLabel, predictions[counter++]);
	}

	return exampleSet;
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
/**
 * This method will apply all the nodes recursively. For each node it will be called when
 * descending the learner hierarchy. The outcomes array stores the information to which node
 * each example of the applySet has been assigned. Each node's model will be applied to the
 * subset of a partitioned example set according to the node's partition id. After the
 * classification has been performed, the examples will be assigned the partion id's of the
 * child nodes, to whose class the examples where classified.
 *
 * It is very important that after each application the predicted label and the confidences are
 * removed explicitly to avoid a memory leak in the memory table!
 *
 * Confidences are multiplied with the outcome every application.
 */
private void performPredictionRecursivly(ExampleSet applySet, Node node, double[] confidences, int[] outcomes,
		int[] depths, int depth, int numberOfPartitions) throws OperatorException {
	if (!node.isLeaf()) {
		// creating partitioned example set
		SplittedExampleSet splittedSet = new SplittedExampleSet(applySet, new Partition(outcomes, numberOfPartitions));
		splittedSet.selectSingleSubset(node.getPartitionId());

		// applying
		ExampleSet currentResultSet = node.getModel().apply(splittedSet);

		// assign each example a child node regarding to the classification outcome
		int resultIndex = 0;
		Attribute predictionAttribute = currentResultSet.getAttributes().getPredictedLabel();
		for (Example example : currentResultSet) {
			int parentIndex = splittedSet.getActualParentIndex(resultIndex);

			// extracting data
			String label = example.getValueAsString(predictionAttribute);
			confidences[parentIndex] *= example.getConfidence(label);

			// setting outcome index according to referenced child node: if child is leaf, this
			// is equivalent to class index
			outcomes[parentIndex] = node.getChild(label).getPartitionId();
			depths[parentIndex] = depth;

			resultIndex++;
		}
		// deleting new columns from example table
		PredictionModel.removePredictedLabel(currentResultSet);

		// now go through children and apply their subset
		for (Node child : node.getChildren()) {
			performPredictionRecursivly(applySet, child, confidences, outcomes, depths, depth + 1, numberOfPartitions);
		}
	}
}
 

import com.rapidminer.operator.learner.PredictionModel; //导入依赖的package包/类
public AbstractStacking(OperatorDescription description, String... subprocessNames) {
	super(description, subprocessNames);
	baseInputExtender.start();
	baseModelExtender.start();
	getTransformer().addRule(baseInputExtender.makePassThroughRule(exampleSetInput));
	getTransformer().addRule(new SubprocessTransformRule(getSubprocess(0)));
	getTransformer().addRule(
			new GeneratePredictionModelTransformationRule(exampleSetInput, modelOutput, PredictionModel.class));
}