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

Example 21 with DerivedFieldMapper

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

the class PMMLNormalizeTranslator method initializeFrom.

/**
 * {@inheritDoc}
 */
@Override
public List<Integer> initializeFrom(final DerivedField[] derivedFields) {
    if (derivedFields == null) {
        return Collections.EMPTY_LIST;
    }
    m_mapper = new DerivedFieldMapper(derivedFields);
    int num = derivedFields.length;
    List<Integer> consumed = new ArrayList<Integer>(num);
    if (num > 0) {
        parseExtensionArray(derivedFields[0].getExtensionArray());
    }
    for (int i = 0; i < derivedFields.length; i++) {
        DerivedField df = derivedFields[i];
        /**
         * This field contains the name of the column in KNIME that
         * corresponds to the derived field in PMML. This is necessary if
         * derived fields are defined on other derived fields and the
         * columns in KNIME are replaced with the preprocessed values.
         * In this case KNIME has to know the original names (e.g. A) while
         * PMML references to A', A'' etc.
         */
        String displayName = df.getDisplayName();
        if (!df.isSetNormContinuous()) {
            // only reading norm continuous other entries are skipped
            continue;
        }
        consumed.add(i);
        NormContinuous normContinuous = df.getNormContinuous();
        if (normContinuous.getLinearNormArray().length > 2) {
            throw new IllegalArgumentException("Only two LinearNorm " + "elements are supported per NormContinuous");
        }
        // String field = normContinuous.getField();
        double[] orig = new double[MAX_NUM_SEGMENTS];
        double[] norm = new double[MAX_NUM_SEGMENTS];
        LinearNorm[] norms = normContinuous.getLinearNormArray();
        for (int j = 0; j < norms.length; j++) {
            orig[j] = norms[j].getOrig();
            norm[j] = norms[j].getNorm();
        }
        double scale = (norm[1] - norm[0]) / (orig[1] - orig[0]);
        m_scales.add(scale);
        m_translations.add(norm[0] - scale * orig[0]);
        if (displayName != null) {
            m_fields.add(displayName);
        } else {
            m_fields.add(m_mapper.getColumnName(normContinuous.getField()));
        }
    }
    return consumed;
}

Example 22 with DerivedFieldMapper

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

the class NumberToStringNodeModel method execute.

/**
 * {@inheritDoc}
 */
@Override
protected PortObject[] execute(final PortObject[] inObjects, final ExecutionContext exec) throws Exception {
    StringBuilder warnings = new StringBuilder();
    BufferedDataTable inData = (BufferedDataTable) inObjects[0];
    DataTableSpec inSpec = inData.getDataTableSpec();
    // find indices to work on.
    List<String> inclcols = m_inclCols.getIncludeList();
    BufferedDataTable resultTable = null;
    if (inclcols.size() == 0) {
        // nothing to convert, let's return the input table.
        resultTable = inData;
        setWarningMessage("No columns selected," + " returning input DataTable.");
    } else {
        int[] indices = findColumnIndices(inData.getSpec());
        ConverterFactory converterFac = new ConverterFactory(indices, inSpec);
        ColumnRearranger colre = new ColumnRearranger(inSpec);
        colre.replace(converterFac, indices);
        resultTable = exec.createColumnRearrangeTable(inData, colre, exec);
        String errorMessage = converterFac.getErrorMessage();
        if (errorMessage.length() > 0) {
            warnings.append("Problems occurred, see Console messages.\n");
        }
        if (warnings.length() > 0) {
            getLogger().warn(errorMessage);
            setWarningMessage(warnings.toString());
        }
    }
    // the optional PMML in port (can be null)
    PMMLPortObject inPMMLPort = m_pmmlInEnabled ? (PMMLPortObject) inObjects[1] : null;
    PMMLStringConversionTranslator trans = new PMMLStringConversionTranslator(m_inclCols.getIncludeList(), StringCell.TYPE, new DerivedFieldMapper(inPMMLPort));
    PMMLPortObjectSpecCreator creator = new PMMLPortObjectSpecCreator(inPMMLPort, inSpec);
    PMMLPortObject outPMMLPort = new PMMLPortObject(creator.createSpec(), inPMMLPort, inSpec);
    outPMMLPort.addGlobalTransformations(trans.exportToTransDict());
    return new PortObject[] { resultTable, outPMMLPort };
}

Example 23 with DerivedFieldMapper

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

the class PMMLGeneralRegressionTranslator method exportTo.

/**
 * {@inheritDoc}
 */
@Override
public SchemaType exportTo(final PMMLDocument pmmlDoc, final PMMLPortObjectSpec spec) {
    m_nameMapper = new DerivedFieldMapper(pmmlDoc);
    GeneralRegressionModel reg = pmmlDoc.getPMML().addNewGeneralRegressionModel();
    final JsonObjectBuilder jsonBuilder = Json.createObjectBuilder();
    if (!m_content.getVectorLengths().isEmpty()) {
        LocalTransformations localTransformations = reg.addNewLocalTransformations();
        for (final Entry<? extends String, ? extends Integer> entry : m_content.getVectorLengths().entrySet()) {
            DataColumnSpec columnSpec = spec.getDataTableSpec().getColumnSpec(entry.getKey());
            if (columnSpec != null) {
                final DataType type = columnSpec.getType();
                final DataColumnProperties props = columnSpec.getProperties();
                final boolean bitVector = type.isCompatible(BitVectorValue.class) || (type.isCompatible(StringValue.class) && props.containsProperty("realType") && "BitVector".equals(props.getProperty("realType")));
                final boolean byteVector = type.isCompatible(ByteVectorValue.class) || (type.isCompatible(StringValue.class) && props.containsProperty("realType") && "ByteVector".equals(props.getProperty("realType")));
                final String lengthAsString;
                final int width;
                if (byteVector) {
                    lengthAsString = "3";
                    width = 4;
                } else if (bitVector) {
                    lengthAsString = "1";
                    width = 1;
                } else {
                    throw new UnsupportedOperationException("Not supported type: " + type + " for column: " + columnSpec);
                }
                for (int i = 0; i < entry.getValue().intValue(); ++i) {
                    final DerivedField derivedField = localTransformations.addNewDerivedField();
                    derivedField.setOptype(OPTYPE.CONTINUOUS);
                    derivedField.setDataType(DATATYPE.INTEGER);
                    derivedField.setName(entry.getKey() + "[" + i + "]");
                    Apply apply = derivedField.addNewApply();
                    apply.setFunction("substring");
                    apply.addNewFieldRef().setField(entry.getKey());
                    Constant from = apply.addNewConstant();
                    from.setDataType(DATATYPE.INTEGER);
                    from.setStringValue(bitVector ? Long.toString(entry.getValue().longValue() - i) : Long.toString(i * width + 1L));
                    Constant length = apply.addNewConstant();
                    length.setDataType(DATATYPE.INTEGER);
                    length.setStringValue(lengthAsString);
                }
            }
            jsonBuilder.add(entry.getKey(), entry.getValue().intValue());
        }
    }
    // PMMLPortObjectSpecCreator newSpecCreator = new PMMLPortObjectSpecCreator(spec);
    // newSpecCreator.addPreprocColNames(m_content.getVectorLengths().entrySet().stream()
    // .flatMap(
    // e -> IntStream.iterate(0, o -> o + 1).limit(e.getValue()).mapToObj(i -> e.getKey() + "[" + i + "]"))
    // .collect(Collectors.toList()));
    PMMLMiningSchemaTranslator.writeMiningSchema(spec, reg);
    // if (!m_content.getVectorLengths().isEmpty()) {
    // Extension miningExtension = reg.getMiningSchema().addNewExtension();
    // miningExtension.setExtender(EXTENDER);
    // miningExtension.setName(VECTOR_COLUMNS_WITH_LENGTH);
    // miningExtension.setValue(jsonBuilder.build().toString());
    // }
    reg.setModelType(getPMMLRegModelType(m_content.getModelType()));
    reg.setFunctionName(getPMMLMiningFunction(m_content.getFunctionName()));
    String algorithmName = m_content.getAlgorithmName();
    if (algorithmName != null && !algorithmName.isEmpty()) {
        reg.setAlgorithmName(algorithmName);
    }
    String modelName = m_content.getModelName();
    if (modelName != null && !modelName.isEmpty()) {
        reg.setModelName(modelName);
    }
    String targetReferenceCategory = m_content.getTargetReferenceCategory();
    if (targetReferenceCategory != null && !targetReferenceCategory.isEmpty()) {
        reg.setTargetReferenceCategory(targetReferenceCategory);
    }
    if (m_content.getOffsetValue() != null) {
        reg.setOffsetValue(m_content.getOffsetValue());
    }
    // add parameter list
    ParameterList paramList = reg.addNewParameterList();
    for (PMMLParameter p : m_content.getParameterList()) {
        Parameter param = paramList.addNewParameter();
        param.setName(p.getName());
        String label = p.getLabel();
        if (label != null) {
            param.setLabel(m_nameMapper.getDerivedFieldName(label));
        }
    }
    // add factor list
    FactorList factorList = reg.addNewFactorList();
    for (PMMLPredictor p : m_content.getFactorList()) {
        Predictor predictor = factorList.addNewPredictor();
        predictor.setName(m_nameMapper.getDerivedFieldName(p.getName()));
    }
    // add covariate list
    CovariateList covariateList = reg.addNewCovariateList();
    for (PMMLPredictor p : m_content.getCovariateList()) {
        Predictor predictor = covariateList.addNewPredictor();
        predictor.setName(m_nameMapper.getDerivedFieldName(p.getName()));
    }
    // add PPMatrix
    PPMatrix ppMatrix = reg.addNewPPMatrix();
    for (PMMLPPCell p : m_content.getPPMatrix()) {
        PPCell cell = ppMatrix.addNewPPCell();
        cell.setValue(p.getValue());
        cell.setPredictorName(m_nameMapper.getDerivedFieldName(p.getPredictorName()));
        cell.setParameterName(p.getParameterName());
        String targetCategory = p.getTargetCategory();
        if (targetCategory != null && !targetCategory.isEmpty()) {
            cell.setTargetCategory(targetCategory);
        }
    }
    // add CovMatrix
    if (m_content.getPCovMatrix().length > 0) {
        PCovMatrix pCovMatrix = reg.addNewPCovMatrix();
        for (PMMLPCovCell p : m_content.getPCovMatrix()) {
            PCovCell covCell = pCovMatrix.addNewPCovCell();
            covCell.setPRow(p.getPRow());
            covCell.setPCol(p.getPCol());
            String tCol = p.getTCol();
            String tRow = p.getTRow();
            if (tRow != null || tCol != null) {
                covCell.setTRow(tRow);
                covCell.setTCol(tCol);
            }
            covCell.setValue(p.getValue());
            String targetCategory = p.getTargetCategory();
            if (targetCategory != null && !targetCategory.isEmpty()) {
                covCell.setTargetCategory(targetCategory);
            }
        }
    }
    // add ParamMatrix
    ParamMatrix paramMatrix = reg.addNewParamMatrix();
    for (PMMLPCell p : m_content.getParamMatrix()) {
        PCell pCell = paramMatrix.addNewPCell();
        String targetCategory = p.getTargetCategory();
        if (targetCategory != null) {
            pCell.setTargetCategory(targetCategory);
        }
        pCell.setParameterName(p.getParameterName());
        pCell.setBeta(p.getBeta());
        Integer df = p.getDf();
        if (df != null) {
            pCell.setDf(BigInteger.valueOf(df));
        }
    }
    return GeneralRegressionModel.type;
}

Example 24 with DerivedFieldMapper

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

the class PMMLSVMTranslator method initializeFrom.

/**
 * {@inheritDoc}
 */
@Override
public void initializeFrom(final PMMLDocument pmmlDoc) {
    m_nameMapper = new DerivedFieldMapper(pmmlDoc);
    SupportVectorMachineModel[] models = pmmlDoc.getPMML().getSupportVectorMachineModelArray();
    if (models.length == 0) {
        throw new IllegalArgumentException("No support vector machine model" + " provided.");
    } else if (models.length > 1) {
        LOGGER.warn("Multiple support vector machine models found. " + "Only the first model is considered.");
    }
    SupportVectorMachineModel svmModel = models[0];
    initKernel(svmModel);
    initSVMs(svmModel);
}

Example 25 with DerivedFieldMapper

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

the class PMMLSVMTranslator method exportTo.

/**
 * {@inheritDoc}
 */
@Override
public SchemaType exportTo(final PMMLDocument pmmlDoc, final PMMLPortObjectSpec spec) {
    m_nameMapper = new DerivedFieldMapper(pmmlDoc);
    PMML pmml = pmmlDoc.getPMML();
    SupportVectorMachineModel svmModel = pmml.addNewSupportVectorMachineModel();
    PMMLMiningSchemaTranslator.writeMiningSchema(spec, svmModel);
    // add support vector machine model attributes
    svmModel.setModelName("SVM");
    svmModel.setFunctionName(MININGFUNCTION.CLASSIFICATION);
    svmModel.setAlgorithmName("Sequential Minimal Optimization (SMO)");
    svmModel.setSvmRepresentation(SVMREPRESENTATION.SUPPORT_VECTORS);
    addTargets(svmModel, spec.getTargetFields().get(0));
    addKernel(svmModel);
    addVectorDictionary(svmModel, spec.getLearningFields());
    addSVMs(svmModel);
    return SupportVectorMachineModel.type;
}