Examples with SparseClassificationTrainingRowBuilder org.knime.base.node.mine.regression.logistic.learner4.data.SparseClassificationTrainingRowBuilder used on opensource projects

Example 1 with SparseClassificationTrainingRowBuilder

use of org.knime.base.node.mine.regression.logistic.learner4.data.SparseClassificationTrainingRowBuilder in project knime-core by knime.

the class LogRegCoordinator method learn.

/**
 * Performs the learning task by creating the appropriate LogRegLearner and all other objects
 * necessary for a successful training.
 *
 * @param trainingData a DataTable that contains the data on which to learn the logistic regression model
 * @param exec the execution context of the corresponding KNIME node
 * @return the content of the logistic regression model
 * @throws InvalidSettingsException if the settings cause inconsistencies during training
 * @throws CanceledExecutionException if the training is canceled
 */
LogisticRegressionContent learn(final BufferedDataTable trainingData, final ExecutionContext exec) throws InvalidSettingsException, CanceledExecutionException {
    CheckUtils.checkArgument(trainingData.size() > 0, "The input table is empty. Please provide data to learn on.");
    CheckUtils.checkArgument(trainingData.size() <= Integer.MAX_VALUE, "The input table contains too many rows.");
    LogRegLearner learner;
    if (m_settings.getSolver() == Solver.IRLS) {
        learner = new IrlsLearner(m_settings.getMaxEpoch(), m_settings.getEpsilon(), m_settings.isCalcCovMatrix());
    } else {
        learner = new SagLogRegLearner(m_settings);
    }
    double calcDomainTime = 1.0 / (5.0 * 2.0 + 1.0);
    exec.setMessage("Analyzing categorical data");
    BufferedDataTable dataTable = recalcDomainForTargetAndLearningFields(trainingData, exec.createSubExecutionContext(calcDomainTime));
    checkConstantLearningFields(dataTable);
    exec.setMessage("Building logistic regression model");
    ExecutionMonitor trainExec = exec.createSubProgress(1.0 - calcDomainTime);
    LogRegLearnerResult result;
    TrainingRowBuilder<ClassificationTrainingRow> rowBuilder = new SparseClassificationTrainingRowBuilder(dataTable, m_pmmlOutSpec, m_settings.getTargetReferenceCategory(), m_settings.getSortTargetCategories(), m_settings.getSortIncludesCategories());
    TrainingData<ClassificationTrainingRow> data;
    Long seed = m_settings.getSeed();
    if (m_settings.isInMemory()) {
        data = new InMemoryData<ClassificationTrainingRow>(dataTable, seed, rowBuilder);
    } else {
        data = new DataTableTrainingData<ClassificationTrainingRow>(trainingData, seed, rowBuilder, m_settings.getChunkSize(), exec.createSilentSubExecutionContext(0.0));
    }
    checkShapeCompatibility(data);
    result = learner.learn(data, trainExec);
    LogisticRegressionContent content = createContentFromLearnerResult(result, rowBuilder, trainingData.getDataTableSpec());
    addToWarning(learner.getWarningMessage());
    return content;
}