sparsegridsalgorithm.java

一个数据挖掘软件ALPHAMINERR的整个过程的JAVA版源代码

/*
 *    This program is free software; you can redistribute it and/or modify
 *    it under the terms of the GNU General Public License as published by
 *    the Free Software Foundation; either version 2 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 General Public License for more details.
 *
 *    You should have received a copy of the GNU General Public License
 *    along with this program; if not, write to the Free Software
 *    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/**
 * Title: XELOPES Data Mining Library
 * Description: The XELOPES library is an open platform-independent and data-source-independent library for Embedded Data Mining.
 * Copyright: Copyright (c) 2002 Prudential Systems Software GmbH
 * Company: ZSoft (www.zsoft.ru), Prudsys (www.prudsys.com)
 * @author Michael Thess
 * @version 1.1
 */
package com.prudsys.pdm.Models.Regression.SparseGrids;

import com.prudsys.pdm.Core.ApplicationAttribute;
import com.prudsys.pdm.Core.MiningException;
import com.prudsys.pdm.Core.MiningModel;
import com.prudsys.pdm.Core.MiningSettings;
import com.prudsys.pdm.Input.MiningArrayStream;
import com.prudsys.pdm.Models.Regression.RegressionAlgorithm;
import com.prudsys.pdm.Models.Supervised.Classifier;
import com.prudsys.pdm.Transform.MiningTransformationActivity;
import com.prudsys.pdm.Transform.Special.ReplaceMissingValueStream;
import com.prudsys.pdm.Transform.Special.TreatOutlierValueStream;

/**
 * A class representing a Sparse Grids algorithm. Each implementation
 * should extend this class and override only the methods:
 * {@link #runAlgorithm() runAlgorithm()},
 * {@link #getClassifier()() getClassifier()}
 */
public abstract class SparseGridsAlgorithm extends RegressionAlgorithm
{
    // -----------------------------------------------------------------------
    //  Variables declarations
    // -----------------------------------------------------------------------
    /** Defines the type of SG (tensor product, simplicial). */
    protected int sgType = SparseGridsSettings.SG_TENSOR_PRODUCT_BASIS_TYPE;

    /** Defines the polynomial degree of SG basis functions. */
    protected int basisDegree = 1;

    /** Is wavelet basis orthogonal? Otherwise, it is just biorthogonal. */
    protected boolean waveletBasis = false;

    /** Include coarse level 0 into calculations? */
    protected boolean coarseGrid = false;

    /** Discretization level. */
    protected int level = 1;

    /** Array of discretization levels if anisotropic grid is used. */
    protected int[] attributeLevels;

    /** Array of all Sparse Grids of the model. */
    protected SparseGrid[] sparseGrids;

    /** Regularization parameter. */
    protected double lambda = 1.0;

    // -----------------------------------------------------------------------
    //  Constructor
    // -----------------------------------------------------------------------
    /**
     * Empty constructor.
     */
    public SparseGridsAlgorithm()
    {
    }

    // -----------------------------------------------------------------------
    //  Getter and setter methods
    // -----------------------------------------------------------------------
    /**
     * Creates an instance of the sparse grids settings class that is required
     * to run the algorithm. The mining settings are assigned through the
     * setMiningSettings method.
     *
     * @return new instance of the sparse grids settings class of the algorithm
     */
    public MiningSettings createMiningSettings() {

      return new SparseGridsSettings();
    }

    /**
     * Set SG settings.
     *
     * @param miningSettings instance of SparseGridsSettings
     * @exception IllegalArgumentException mining settings are not of sparse grids type
     */
    public final void setMiningSettings( MiningSettings miningSettings ) throws IllegalArgumentException
    {
        if ( miningSettings instanceof SparseGridsSettings )
        {
            super.setMiningSettings( miningSettings );
            SparseGridsSettings sgs = (SparseGridsSettings) miningSettings;
            this.sgType          = sgs.getSgType();
            this.basisDegree     = sgs.getBasisDegree();
            this.waveletBasis    = sgs.isWaveletBasis();
            this.coarseGrid      = sgs.isCoarseGrid();
            this.level           = sgs.getLevel();
            this.attributeLevels = sgs.getAttributeLevels();
            this.lambda          = sgs.getLambda();
        }
        else
        {
            throw new IllegalArgumentException( "MiningSettings have to be SparseGridsSettings." );
        };
    }

    /**
     * Returns SG classifier.
     *
     * @return SG classifier
     */
    public abstract Classifier getClassifier();

    // -----------------------------------------------------------------------
    //  Run SG algorithm and build mining model
    // -----------------------------------------------------------------------
    /**
     * Runs SG algorithm.
     *
     * @exception MiningException could not run algorithm
     */
    protected abstract void runAlgorithm() throws MiningException;

    /**
     * Builds mining model by running the SG algorithm internally.
     * Before starting the algorithm, missing values are replaced.
     *
     * @return sparse grids mining model generated by the algorithm
     * @exception MiningException could not build model
     */
    public MiningModel buildModel() throws MiningException
    {
      long start = ( new java.util.Date() ).getTime();

      // Outlier treatment and missing value replacement:
      TreatOutlierValueStream tro   = new TreatOutlierValueStream(miningInputStream);
      tro.setNumOutliers( ApplicationAttribute.OUTLIER_TREATMENT_METHOD_asExtremeValues );
      tro.createTreatOutlierValueTransformationStep();

      ReplaceMissingValueStream rep = new ReplaceMissingValueStream(miningInputStream);
      miningInputStream             = new MiningArrayStream( rep.createReplaceMissingValueStream() );

      // Run SG algorithm:
      runAlgorithm();

      // Build SG model:
      SparseGridsMiningModel model = new SparseGridsMiningModel();
      model.setMiningSettings( miningSettings );
      model.setInputSpec( applicationInputSpecification );
      model.setTarget( applicationInputSpecification.getTargetApplicationAttribute() );

      // Outlier treatment and missing value in application input specification:
      // Create inner transformation object:
      MiningTransformationActivity mta = new MiningTransformationActivity();
      mta.addTransformationStep( tro.getMts() );
      mta.addTransformationStep( rep.getMts() );
      model.setMiningTransform( mta );

      // Outliers and missing values in application input specification:
      applicationInputSpecification.setInputSpecFromInnerTrafo(metaData, tro, rep);

      // Set SG parameter:
      model.setSgType(sgType);
      model.setBasisDegree(basisDegree);
      model.setWaveletBasis(waveletBasis);
      model.setCoarseGrid(coarseGrid);
      model.setLevel(level);
      model.setAttributeLevels(attributeLevels);

      // Set classifier:
      model.setClassifier( getClassifier() );

      this.miningModel = model;

      long end = ( new java.util.Date() ).getTime();
      timeSpentToBuildModel = ( end - start ) / 1000.0;

      return model;
    }
}