Examples with PMMLDocument org.dmg.pmml.PMMLDocument used on opensource projects

Example 11 with PMMLDocument

use of org.dmg.pmml.PMMLDocument in project knime-core by knime.

the class PMMLTreeModelWrapper method createPMMLDocument.

/**
 * {@inheritDoc}
 */
@Override
public PMMLDocument createPMMLDocument(final DataDictionary dataDict) {
    PMMLDocument doc = createEmptyDocument(dataDict);
    doc.getPMML().setTreeModelArray(new TreeModel[] { m_model });
    return doc;
}

Example 12 with PMMLDocument

use of org.dmg.pmml.PMMLDocument in project knime-core by knime.

the class PMMLRuleSetModelWrapper method createPMMLDocument.

/**
 * {@inheritDoc}
 */
@Override
public PMMLDocument createPMMLDocument(final DataDictionary dataDict) {
    PMMLDocument doc = createEmptyDocument(dataDict);
    doc.getPMML().setRuleSetModelArray(new RuleSetModel[] { m_model });
    return doc;
}

Example 13 with PMMLDocument

use of org.dmg.pmml.PMMLDocument in project knime-core by knime.

the class PMMLKNNTranslator method exportTo.

/**
 * {@inheritDoc}
 */
@Override
public SchemaType exportTo(final PMMLDocument pmmlDoc, final PMMLPortObjectSpec spec) {
    LinkedHashMap<Integer, String> columnNames = new LinkedHashMap<Integer, String>();
    DataTableSpec tSpec = m_table.getDataTableSpec();
    // Find learning columns and store them in the map for later
    for (String lc : m_includes) {
        columnNames.put(tSpec.findColumnIndex(lc), "col" + columnNames.size());
    }
    // Create initial XML elements
    PMML pmml = pmmlDoc.getPMML();
    NearestNeighborModel knn = pmml.addNewNearestNeighborModel();
    PMMLMiningSchemaTranslator.writeMiningSchema(spec, knn);
    knn.setAlgorithmName("K-Nearest Neighbors");
    knn.setFunctionName(org.dmg.pmml.MININGFUNCTION.CLASSIFICATION);
    knn.setNumberOfNeighbors(BigInteger.valueOf(m_numNeighbors));
    // Only euclidean is supported so far
    ComparisonMeasure cm = knn.addNewComparisonMeasure();
    cm.addNewEuclidean();
    // KNNInputs is a list of the fields used for determining the distance
    KNNInputs inputs = knn.addNewKNNInputs();
    for (int i : columnNames.keySet()) {
        KNNInput input = inputs.addNewKNNInput();
        String col = tSpec.getColumnSpec(i).getName();
        input.setField(col);
        input.setCompareFunction(COMPAREFUNCTION.ABS_DIFF);
    }
    TrainingInstances ti = knn.addNewTrainingInstances();
    // Here we create a mapping from column name to name of the XML element for the column's values
    InstanceFields instanceFields = ti.addNewInstanceFields();
    for (int i : columnNames.keySet()) {
        InstanceField instanceField = instanceFields.addNewInstanceField();
        String col = tSpec.getColumnSpec(i).getName();
        instanceField.setField(col);
        instanceField.setColumn(columnNames.get(i));
    }
    int targetIdx = tSpec.findColumnIndex(spec.getTargetFields().get(0));
    InstanceField target = instanceFields.addNewInstanceField();
    target.setField(spec.getTargetFields().get(0));
    target.setColumn("target");
    // The inline table holds the actual data.
    // We use the map we created in the beginning to determine the element xml-element-names
    InlineTable it = ti.addNewInlineTable();
    Document doc = it.getDomNode().getOwnerDocument();
    int counter = 0;
    for (DataRow row : m_table) {
        // Stop if we have reached the maximum number of records
        if (m_maxRecords > -1 && ++counter > m_maxRecords) {
            break;
        }
        Row inlineRow = it.addNewRow();
        Element rowNode = (Element) inlineRow.getDomNode();
        for (int col : columnNames.keySet()) {
            Element field = doc.createElementNS(PMMLUtils.getPMMLCurrentVersionNamespace(), columnNames.get(col));
            field.appendChild(doc.createTextNode(row.getCell(col).toString()));
            rowNode.appendChild(field);
        }
        Element targetField = doc.createElementNS(PMMLUtils.getPMMLCurrentVersionNamespace(), "target");
        targetField.appendChild(doc.createTextNode(row.getCell(targetIdx).toString()));
        rowNode.appendChild(targetField);
    }
    return NearestNeighborModel.type;
}

Example 14 with PMMLDocument

use of org.dmg.pmml.PMMLDocument in project knime-core by knime.

the class XML2PMMLNodeModel method createColRearranger.

private ColumnRearranger createColRearranger(final DataTableSpec spec) throws InvalidSettingsException {
    if (m_xmlColumnName.getStringValue() == null) {
        guessDefaultXMLColumn(spec);
    }
    String xmlColName = m_xmlColumnName.getStringValue();
    String newColName = m_newColumnName.getStringValue();
    final int colIndex = spec.findColumnIndex(xmlColName);
    CheckUtils.checkSetting(colIndex >= 0, "Column: '%s' does not exist anymore.", xmlColName);
    final DataColumnSpec colSpec = spec.getColumnSpec(colIndex);
    CheckUtils.checkSetting(colSpec.getType().isCompatible(StringValue.class), "Selected column '%s' is not string/xml-compatible", xmlColName);
    DataColumnSpecCreator colSpecCreator;
    if (newColName != null && !m_replaceColumn.getBooleanValue()) {
        String newName = DataTableSpec.getUniqueColumnName(spec, newColName);
        colSpecCreator = new DataColumnSpecCreator(newName, PMMLCell.TYPE);
    } else {
        colSpecCreator = new DataColumnSpecCreator(colSpec);
        colSpecCreator.setType(PMMLCell.TYPE);
        colSpecCreator.removeAllHandlers();
        colSpecCreator.setDomain(null);
    }
    DataColumnSpec outColumnSpec = colSpecCreator.createSpec();
    ColumnRearranger rearranger = new ColumnRearranger(spec);
    CellFactory fac = new SingleCellFactory(outColumnSpec) {

        @Override
        public DataCell getCell(final DataRow row) {
            DataCell cell = row.getCell(colIndex);
            if (cell.isMissing()) {
                return DataType.getMissingCell();
            } else {
                PMMLDocument pmmlDoc = null;
                String failure = null;
                XmlObject xmlDoc;
                try (LockedSupplier<Document> supplier = ((XMLValue<Document>) cell).getDocumentSupplier()) {
                    xmlDoc = XmlObject.Factory.parse(supplier.get().cloneNode(true));
                    if (xmlDoc instanceof PMMLDocument) {
                        pmmlDoc = (PMMLDocument) xmlDoc;
                    } else if (PMMLUtils.isOldKNIMEPMML(xmlDoc) || PMMLUtils.is4_1PMML(xmlDoc)) {
                        String updatedPMML = PMMLUtils.getUpdatedVersionAndNamespace(xmlDoc);
                        /* Parse the modified document and assign it to a
                                 * PMMLDocument.*/
                        pmmlDoc = PMMLDocument.Factory.parse(updatedPMML);
                    } else {
                        failure = "No valid PMML v 3.x/4.0/4.1 document";
                    }
                } catch (XmlException e) {
                    if (!m_failOnInvalid.getBooleanValue()) {
                        LOGGER.error("Invalid PMML in row " + row.getKey() + ": " + e.getMessage(), e);
                    }
                    failure = e.getMessage();
                }
                if (failure != null) {
                    m_failCounter.incrementAndGet();
                    if (m_failOnInvalid.getBooleanValue()) {
                        throw new RuntimeException("Invalid PMML in row " + row.getKey() + ": " + failure);
                    } else {
                        return new MissingCell(failure);
                    }
                } else {
                    try {
                        return PMMLCellFactory.create(pmmlDoc.toString());
                    } catch (Exception e) {
                        return new MissingCell(e.getMessage());
                    }
                }
            }
        }
    };
    if (m_replaceColumn.getBooleanValue()) {
        rearranger.replace(fac, colIndex);
    } else {
        rearranger.append(fac);
    }
    return rearranger;
}

Example 15 with PMMLDocument

use of org.dmg.pmml.PMMLDocument in project knime-core by knime.

the class PMMLNaiveBayesModelWrapper method createPMMLDocument.

/**
 * {@inheritDoc}
 */
@Override
public PMMLDocument createPMMLDocument(final DataDictionary dataDict) {
    PMMLDocument doc = createEmptyDocument(dataDict);
    doc.getPMML().setNaiveBayesModelArray(new NaiveBayesModel[] { m_model });
    return doc;
}