package de.lmu.ifi.dbs.elki.algorithm.clustering.correlation;
/*
This file is part of ELKI:
Environment for Developing KDD-Applications Supported by Index-Structures
Copyright (C) 2012
Ludwig-Maximilians-Universität München
Lehr- und Forschungseinheit für Datenbanksysteme
ELKI Development Team
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
import org.junit.Test;
import de.lmu.ifi.dbs.elki.JUnit4Test;
import de.lmu.ifi.dbs.elki.algorithm.AbstractSimpleAlgorithmTest;
import de.lmu.ifi.dbs.elki.algorithm.clustering.DBSCAN;
import de.lmu.ifi.dbs.elki.data.Clustering;
import de.lmu.ifi.dbs.elki.data.DoubleVector;
import de.lmu.ifi.dbs.elki.data.model.CorrelationModel;
import de.lmu.ifi.dbs.elki.database.Database;
import de.lmu.ifi.dbs.elki.distance.distancefunction.correlation.ERiCDistanceFunction;
import de.lmu.ifi.dbs.elki.index.preprocessed.localpca.KNNQueryFilteredPCAIndex;
import de.lmu.ifi.dbs.elki.math.linearalgebra.pca.PCAFilteredRunner;
import de.lmu.ifi.dbs.elki.math.linearalgebra.pca.PCARunner;
import de.lmu.ifi.dbs.elki.math.linearalgebra.pca.PercentageEigenPairFilter;
import de.lmu.ifi.dbs.elki.math.linearalgebra.pca.RelativeEigenPairFilter;
import de.lmu.ifi.dbs.elki.math.linearalgebra.pca.WeightedCovarianceMatrixBuilder;
import de.lmu.ifi.dbs.elki.math.linearalgebra.pca.weightfunctions.ErfcWeight;
import de.lmu.ifi.dbs.elki.utilities.ClassGenericsUtil;
import de.lmu.ifi.dbs.elki.utilities.optionhandling.ParameterException;
import de.lmu.ifi.dbs.elki.utilities.optionhandling.parameterization.ListParameterization;
/**
* Perform a full ERiC run, and compare the result with a clustering derived
* from the data set labels. This test ensures that ERiC performance doesn't
* unexpectedly drop on this data set (and also ensures that the algorithms
* work, as a side effect).
*
* @author Erich Schubert
* @author Katharina Rausch
*/
public class TestERiCResults extends AbstractSimpleAlgorithmTest implements JUnit4Test {
/**
* Run ERiC with fixed parameters and compare the result to a golden standard.
*
* @throws ParameterException on errors.
*/
@Test
public void testERiCResults() {
Database db = makeSimpleDatabase(UNITTEST + "hierarchical-3d2d1d.csv", 600);
// ERiC
ListParameterization params = new ListParameterization();
params.addParameter(COPAC.PARTITION_ALGORITHM_ID, DBSCAN.class);
params.addParameter(DBSCAN.MINPTS_ID, 30);
params.addParameter(DBSCAN.EPSILON_ID, 0);
// ERiC Distance function in DBSCAN:
params.addParameter(COPAC.PARTITION_DISTANCE_ID, ERiCDistanceFunction.class);
params.addParameter(ERiCDistanceFunction.DELTA_ID, 0.20);
params.addParameter(ERiCDistanceFunction.TAU_ID, 0.04);
// Preprocessing via Local PCA:
params.addParameter(COPAC.PREPROCESSOR_ID, KNNQueryFilteredPCAIndex.Factory.class);
params.addParameter(KNNQueryFilteredPCAIndex.Factory.K_ID, 50);
// PCA
params.addParameter(PCARunner.PCA_COVARIANCE_MATRIX, WeightedCovarianceMatrixBuilder.class);
params.addParameter(WeightedCovarianceMatrixBuilder.WEIGHT_ID, ErfcWeight.class);
params.addParameter(PCAFilteredRunner.PCA_EIGENPAIR_FILTER, RelativeEigenPairFilter.class);
params.addParameter(RelativeEigenPairFilter.EIGENPAIR_FILTER_RALPHA, 1.60);
ERiC<DoubleVector> eric = ClassGenericsUtil.parameterizeOrAbort(ERiC.class, params);
testParameterizationOk(params);
// run ERiC on database
Clustering<CorrelationModel<DoubleVector>> result = eric.run(db);
testFMeasure(db, result, 0.714207); // Hierarchical pairs scored: 0.9204825
testClusterSizes(result, new int[] { 109, 184, 307 });
}
/**
* Run ERiC with fixed parameters and compare the result to a golden standard.
*
* @throws ParameterException on errors.
*/
@Test
public void testERiCOverlap() {
Database db = makeSimpleDatabase(UNITTEST + "correlation-overlap-3-5d.ascii", 650);
// Setup algorithm
ListParameterization params = new ListParameterization();
// ERiC
params.addParameter(COPAC.PARTITION_ALGORITHM_ID, DBSCAN.class);
params.addParameter(DBSCAN.MINPTS_ID, 15);
params.addParameter(DBSCAN.EPSILON_ID, 0);
// ERiC Distance function in DBSCAN:
params.addParameter(COPAC.PARTITION_DISTANCE_ID, ERiCDistanceFunction.class);
params.addParameter(ERiCDistanceFunction.DELTA_ID, 1.0);
params.addParameter(ERiCDistanceFunction.TAU_ID, 1.0);
// Preprocessing via Local PCA:
params.addParameter(COPAC.PREPROCESSOR_ID, KNNQueryFilteredPCAIndex.Factory.class);
params.addParameter(KNNQueryFilteredPCAIndex.Factory.K_ID, 45);
// PCA
params.addParameter(PCARunner.PCA_COVARIANCE_MATRIX, WeightedCovarianceMatrixBuilder.class);
params.addParameter(WeightedCovarianceMatrixBuilder.WEIGHT_ID, ErfcWeight.class);
params.addParameter(PCAFilteredRunner.PCA_EIGENPAIR_FILTER, PercentageEigenPairFilter.class);
params.addParameter(PercentageEigenPairFilter.ALPHA_ID, 0.6);
ERiC<DoubleVector> eric = ClassGenericsUtil.parameterizeOrAbort(ERiC.class, params);
testParameterizationOk(params);
// run ERiC on database
Clustering<CorrelationModel<DoubleVector>> result = eric.run(db);
testFMeasure(db, result, 0.831136946);
testClusterSizes(result, new int[] { 29, 189, 207, 225 });
}
}