supportvectoralgorithm.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.SVM;

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.Input.MiningFilterStream;
import com.prudsys.pdm.Input.MiningInputStream;
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.MiningTransformationStep;
import com.prudsys.pdm.Transform.Special.ReplaceMissingValueStream;

/**
 * A class representing a Support Vector Machine algorithm. Each implementation
 * should extend this class and override only the methods:
 * {@link #runAlgorithm() runAlgorithm()},
 * {@link #getClassifier()() getClassifier()}
 */
public abstract class SupportVectorAlgorithm extends RegressionAlgorithm
{
    // -----------------------------------------------------------------------
    //  Variables declarations
    // -----------------------------------------------------------------------
    // SVM types:
    protected int svmType    = SupportVectorSettings.SVM_C_SVC;
    protected int kernelType = SupportVectorSettings.KERNEL_RBF;

    // Kernel parameters:
    protected double degree = 3.0;
    protected double gamma  = 1.0;
    protected double coef0  = 0.0;

    // Algorithm parameters:
    protected double C           = 1.0;
    protected double nu          = 0.5;
    protected double lossEpsilon = 0.1;

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

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

      return new SupportVectorSettings();
    }

    /**
     * Set SVM settings.
     *
     * @param miningSettings instance of SupportVectorSettings
     * @exception IllegalArgumentException mining settings are not of support vector type
     */
    public final void setMiningSettings( MiningSettings miningSettings ) throws IllegalArgumentException
    {
        if ( miningSettings instanceof SupportVectorSettings )
        {
            super.setMiningSettings( miningSettings );
            SupportVectorSettings svs = (SupportVectorSettings) miningSettings;
            this.svmType     = svs.getSvmType();
            this.kernelType  = svs.getKernelType();
            this.degree      = svs.getDegree();
            this.gamma       = svs.getGamma();
            this.coef0       = svs.getCoef0();
            this.C           = svs.getC();
            this.nu          = svs.getNu();
            this.lossEpsilon = svs.getLossEpsilon();
        }
        else
        {
            throw new IllegalArgumentException( "MiningSettings have to be SupportVectorSettings." );
        };
    }

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

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

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

      // Create inner trafo object:
      MiningTransformationActivity innerTrafo = new MiningTransformationActivity();

      // Replace missing values trafo:
      ReplaceMissingValueStream rep = new ReplaceMissingValueStream(miningInputStream);
      MiningInputStream repStream   = rep.createReplaceMissingValueStream();
      innerTrafo.addTransformationStep( rep.getMts() );

      // Remove supplementary attributes trafo:
      MiningInputStream mis = repStream;
      MiningTransformationStep removeTrafo = applicationInputSpecification.createRemoveTrafoFromInputSpec();
      if (removeTrafo != null) {
        mis = new MiningFilterStream(repStream, removeTrafo);
        innerTrafo.addTransformationStep(removeTrafo);
      }

      // Copy stream to array:
      miningInputStream = new MiningArrayStream(mis);

      // Run SVM algorithm:
      runAlgorithm();

      // Build SVM model:
      SupportVectorMiningModel model = new SupportVectorMiningModel();
      model.setMiningSettings( miningSettings );
      model.setInputSpec( applicationInputSpecification );
      model.setTarget( applicationInputSpecification.getTargetApplicationAttribute() );
      model.setMiningTransform( innerTrafo );

      // Missing values in application input specification:
      applicationInputSpecification.setInputSpecFromInnerTrafo(metaData, null, rep);

      // Set SVM parameter:
      model.setSvmType(svmType);
      model.setKernelType(kernelType);
      model.setDegree(degree);
      model.setCoef0(coef0);
      model.setGamma(gamma);

      // Set classifier:
      SupportVectorClassifier svmClassifier = (SupportVectorClassifier) getClassifier();
      model.setClassifier( svmClassifier );
      model.setSupportVectors( svmClassifier.getSupportVectors() );
      model.setCoefficients( svmClassifier.getCoefficients() );
      model.setAbsoluteCoefficient( svmClassifier.getAbsoluteCoefficient() );

      this.miningModel = model;

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

      return model;
    }
}