Examples with PMMLPortObjectSpec org.knime.core.node.port.pmml.PMMLPortObjectSpec used on opensource projects

Example 21 with PMMLPortObjectSpec

use of org.knime.core.node.port.pmml.PMMLPortObjectSpec in project knime-core by knime.

the class LearnerTest method testPerformChdAgeData.

/**
 * Test method for {@link org.knime.base.node.mine.regression.logistic.learner.Learner#perform(BufferedDataTable, org.knime.core.node.ExecutionContext)}.
 * @throws CanceledExecutionException
 */
@Test
public final void testPerformChdAgeData() throws Exception {
    final BufferedDataTable data = m_exec.createBufferedDataTable(new ChdAgeData(), m_exec);
    PMMLPortObjectSpecCreator specCreator = new PMMLPortObjectSpecCreator(data.getDataTableSpec());
    specCreator.setLearningColsNames(Arrays.asList(new String[] { "Age" }));
    specCreator.setTargetColName("Evidence of Coronary Heart Disease");
    final PMMLPortObjectSpec spec = specCreator.createSpec();
    // done in KNIME thread pool, expected by code
    Future<LogisticRegressionContent> callable = KNIMEConstants.GLOBAL_THREAD_POOL.enqueue(new Callable<LogisticRegressionContent>() {

        @Override
        public LogisticRegressionContent call() throws Exception {
            final Learner learner = new Learner(spec, null, true, true);
            return learner.perform(data, m_exec);
        }
    });
    LogisticRegressionContent content = callable.get();
    // Reference results are published in the book:
    // Applied Logistic Regression,
    // David W. Hosmer and Stanley Lemeshow
    // Wiley, 2000 (2nd. ed)
    // The table of results are found on page 10
    Assert.assertEquals(-53.67656, content.getEstimatedLikelihood(), 0.001);
}

Example 22 with PMMLPortObjectSpec

use of org.knime.core.node.port.pmml.PMMLPortObjectSpec in project knime-core by knime.

the class LogisticRegressionContent method load.

/**
 * @param parContent the content that holds the internals
 * @param spec the data table spec of the training data
 * @return a instance with he loaded values
 * @throws InvalidSettingsException when data are not well formed
 */
static LogisticRegressionContent load(final ModelContentRO parContent, final DataTableSpec spec) throws InvalidSettingsException {
    String target = parContent.getString(CFG_TARGET);
    String[] learningCols = parContent.getStringArray(CFG_LEARNING_COLS);
    PMMLPortObjectSpec pmmlSpec = createSpec(spec, target, learningCols);
    String[] factors = parContent.getStringArray(CFG_FACTORS);
    // Since 3.0
    final String[] vectorColumns = parContent.getStringArray(CFG_VECTOR_NAMES);
    final int[] vectorLengths = parContent.getIntArray(CFG_VECTOR_LENGTHS);
    CheckUtils.checkSetting(vectorColumns.length == vectorLengths.length, "The length of vector names and their length do not match: " + vectorColumns.length + " <> " + vectorLengths.length);
    Map<String, Integer> vectorLengthMap = new LinkedHashMap<>();
    for (int i = 0; i < vectorLengths.length; i++) {
        vectorLengthMap.put(vectorColumns[i], vectorLengths[i]);
    }
    String[] covariates = parContent.getStringArray(CFG_COVARIATES);
    double[] coeff = parContent.getDoubleArray(CFG_COEFFICIENTS);
    double likelihood = parContent.getDouble(CFG_LOG_LIKELIHOOD);
    RealMatrix covMat = null;
    if (parContent.getBoolean(CFG_COVMAT_PRESENT)) {
        covMat = toMatrix(parContent.getDoubleArray(CFG_COVARIANCE_MATRIX), coeff.length);
    }
    int iter = parContent.getInt(CFG_ITER);
    // introduced in 2.9
    DataCell targetReferenceCategory = parContent.getDataCell(CFG_TARGET_REFERENCE_CATEGORY, null);
    boolean sortTargetCategories = parContent.getBoolean(CFG_SORT_TARGET_CATEGORIES, true);
    boolean sortFactorsCategories = parContent.getBoolean(CFG_SORT_FACTORS_CATEGORIES, true);
    return new LogisticRegressionContent(pmmlSpec, Arrays.asList(factors), Arrays.asList(covariates), vectorLengthMap, targetReferenceCategory, sortTargetCategories, sortFactorsCategories, toMatrix(coeff, coeff.length), likelihood, covMat, iter);
}

Example 23 with PMMLPortObjectSpec

use of org.knime.core.node.port.pmml.PMMLPortObjectSpec in project knime-core by knime.

the class SVMPredictorNodeModel method createColumnRearranger.

private ColumnRearranger createColumnRearranger(final PMMLPortObject pmmlModel, final DataTableSpec inSpec) throws InvalidSettingsException {
    List<Node> models = pmmlModel.getPMMLValue().getModels(PMMLModelType.SupportVectorMachineModel);
    if (models.isEmpty()) {
        String msg = "SVM evaluation failed: " + "No support vector machine model found.";
        LOGGER.error(msg);
        throw new RuntimeException(msg);
    }
    PMMLSVMTranslator trans = new PMMLSVMTranslator();
    pmmlModel.initializeModelTranslator(trans);
    List<Svm> svms = trans.getSVMs();
    m_svms = svms.toArray(new Svm[svms.size()]);
    if (m_addProbabilities.getBooleanValue() == pmmlModel.getSpec().getTargetCols().size() > 0) {
        adjustOrder(pmmlModel.getSpec().getTargetCols().get(0));
    }
    DataTableSpec testSpec = inSpec;
    PMMLPortObjectSpec pmmlSpec = pmmlModel.getSpec();
    DataTableSpec trainingSpec = pmmlSpec.getDataTableSpec();
    // try to find all columns (except the class column)
    Vector<Integer> colindices = new Vector<Integer>();
    for (DataColumnSpec colspec : trainingSpec) {
        if (colspec.getType().isCompatible(DoubleValue.class)) {
            int colindex = testSpec.findColumnIndex(colspec.getName());
            if (colindex < 0) {
                throw new InvalidSettingsException("Column " + "\'" + colspec.getName() + "\' not found" + " in test data");
            }
            colindices.add(colindex);
        }
    }
    m_colindices = new int[colindices.size()];
    for (int i = 0; i < m_colindices.length; i++) {
        m_colindices[i] = colindices.get(i);
    }
    final PredictorHelper predictorHelper = PredictorHelper.getInstance();
    final String targetCol = pmmlSpec.getTargetFields().iterator().next();
    SVMPredictor svmpredict = new SVMPredictor(targetCol, m_svms, m_colindices, predictorHelper.computePredictionColumnName(m_predictionColumn.getStringValue(), m_overridePrediction.getBooleanValue(), targetCol), m_addProbabilities.getBooleanValue(), m_suffix.getStringValue());
    ColumnRearranger colre = new ColumnRearranger(testSpec);
    colre.append(svmpredict);
    return colre;
}

Example 24 with PMMLPortObjectSpec

use of org.knime.core.node.port.pmml.PMMLPortObjectSpec in project knime-core by knime.

the class PredictorHelper method createOutTableSpec.

/**
 * Computes the output table's specifaction based on common node settings.
 *
 * @param dataSpec The input table {@link DataColumnSpec}.
 * @param modelSpec The model {@link PMMLPortObjectSpec}.
 * @param addProbs Add the probability columns?
 * @param predictionCol Custom name of the prediction column.
 * @param shouldOverride Should we use that name?
 * @param suffix Suffix for probability columns.
 * @return The output table {@link DataTableSpec}.
 * @throws InvalidSettingsException Invalid settings for the prediction column name.
 */
public DataTableSpec createOutTableSpec(final PortObjectSpec dataSpec, final PortObjectSpec modelSpec, final boolean addProbs, final String predictionCol, final boolean shouldOverride, final String suffix) throws InvalidSettingsException {
    CheckUtils.checkSettingNotNull(predictionCol, "Prediction column name cannot be null");
    CheckUtils.checkSetting(!predictionCol.trim().isEmpty(), "Prediction column name cannot be empty");
    List<DataCell> predValues = null;
    if (addProbs) {
        predValues = getPredictionValues((PMMLPortObjectSpec) modelSpec);
        if (predValues == null) {
            // no out spec can be determined
            return null;
        }
    }
    int numCols = (predValues == null ? 0 : predValues.size()) + 1;
    DataTableSpec inSpec = (DataTableSpec) dataSpec;
    DataColumnSpec[] newCols = new DataColumnSpec[numCols];
    /* Set bar renderer and domain [0,1] as default for the double cells
         * containing the distribution */
    // DataColumnProperties propsRendering = new DataColumnProperties(
    // Collections.singletonMap(
    // DataValueRenderer.PROPERTY_PREFERRED_RENDERER,
    // DoubleBarRenderer.DESCRIPTION));
    DataColumnDomain domain = new DataColumnDomainCreator(new DoubleCell(0.0), new DoubleCell(1.0)).createDomain();
    String trainingColumnName = ((PMMLPortObjectSpec) modelSpec).getTargetFields().iterator().next();
    // add all distribution columns
    for (int i = 0; i < numCols - 1; i++) {
        assert predValues != null;
        DataColumnSpecCreator colSpecCreator = new DataColumnSpecCreator(probabilityColumnName(trainingColumnName, predValues.get(i).toString(), suffix), DoubleCell.TYPE);
        // colSpecCreator.setProperties(propsRendering);
        colSpecCreator.setDomain(domain);
        newCols[i] = colSpecCreator.createSpec();
    }
    // add the prediction column
    String predictionColumnName = computePredictionColumnName(predictionCol, shouldOverride, trainingColumnName);
    newCols[numCols - 1] = new DataColumnSpecCreator(predictionColumnName, StringCell.TYPE).createSpec();
    DataTableSpec newColSpec = new DataTableSpec(newCols);
    return new DataTableSpec(inSpec, newColSpec);
}

Example 25 with PMMLPortObjectSpec

use of org.knime.core.node.port.pmml.PMMLPortObjectSpec in project knime-core by knime.

the class RegressionPredictorNodeDialogPane method loadSettingsFrom.

/**
 * {@inheritDoc}
 */
@Override
protected void loadSettingsFrom(final NodeSettingsRO settings, final PortObjectSpec[] specs) throws NotConfigurableException {
    RegressionPredictorSettings s = new RegressionPredictorSettings(m_showProbOptions);
    s.loadSettingsForDialog(settings);
    m_hasCustomPredictionName.setSelected(s.getHasCustomPredictionName());
    PMMLPortObjectSpec portSpec = (PMMLPortObjectSpec) specs[0];
    DataTableSpec tableSpec = (DataTableSpec) specs[1];
    // check if the model has the correct target type
    try {
        RegressionPredictorNodeModel.checkModelTargetType(portSpec, m_showProbOptions);
    } catch (InvalidSettingsException e1) {
        throw new NotConfigurableException(e1.getMessage());
    }
    if (s.getCustomPredictionName() != null) {
        m_customPredictionName.setText(s.getCustomPredictionName());
    } else {
        try {
            DataColumnSpec[] outSpec = RegressionPredictorCellFactory.createColumnSpec(portSpec, tableSpec, new RegressionPredictorSettings(m_showProbOptions));
            m_customPredictionName.setText(outSpec[outSpec.length - 1].getName());
        } catch (InvalidSettingsException e) {
        // Open dialog and give a chance define settings
        }
    }
    if (m_showProbOptions) {
        m_includeProbs.setSelected(s.getIncludeProbabilities());
        m_probColumnSuffix.setText(s.getPropColumnSuffix());
    }
    updateEnableState();
}