svmtimeseriesalgorithm.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
 * @author Victor Borichev
 * @author Valentine Stepanenko ([e-mail])
 * @version 1.0
 */

package com.prudsys.pdm.Models.TimeSeriesPredict.Algorithms.Multidimensional;

import com.prudsys.pdm.Core.MiningAlgorithm;
import com.prudsys.pdm.Core.MiningAlgorithmSpecification;
import com.prudsys.pdm.Core.MiningAttribute;
import com.prudsys.pdm.Core.MiningException;
import com.prudsys.pdm.Input.MiningStoredData;
import com.prudsys.pdm.Models.Regression.RegressionDeviationAssessment;
import com.prudsys.pdm.Models.Regression.SVM.SupportVectorMiningModel;
import com.prudsys.pdm.Models.Regression.SVM.SupportVectorSettings;
import com.prudsys.pdm.Models.Supervised.SupervisedMiningModel;
import com.prudsys.pdm.Models.TimeSeriesPredict.TimeSeriesMiningAlgorithm;
import com.prudsys.pdm.Transform.Special.BlockVectorStream;
import com.prudsys.pdm.Utils.GeneralUtils;

/**
 * Implementation of multidimensional SVM-based time series prognosis
 * using regression methods.
 */
public class SVMTimeSeriesAlgorithm extends TimeSeriesMiningAlgorithm {

  SupportVectorMiningModel[] svmm;

  String regressionMethod;

  /**
   * Empty constructor.
   */
  public SVMTimeSeriesAlgorithm() {

  }

  /**
   * Run algorithm.
   *
   * @exception MiningException could not create time series models
   */
  protected void runAlgorithm() throws MiningException {

    System.out.print("Building classifiers...");
    BlockVectorStream bvStream = new BlockVectorStream();
    bvStream.setBlockLength( embeddingDimension );
    bvStream.setStepSize( stepSize );

    int nAtt = miningInputStream.getMetaData().getAttributesNumber();
    svmm = new SupportVectorMiningModel[nAtt];
    for (int i = 0; i < nAtt; i++) {

      // Create mining input stream for training i-th classifie:
      MiningStoredData msd = new MiningStoredData();
      MiningAttribute tarAtt = miningInputStream.getMetaData().getMiningAttribute(i);
      miningInputStream.reset();
      bvStream.transformTarget(miningInputStream, tarAtt, msd);

      // Display transformed stream:
      System.out.println("Transformed: ");
        System.out.println("nAtt = "+ msd.getMetaData().getAttributesNumber() +
                           " nVec = " + msd.getVectorsNumber() );
      System.out.println(msd);

      // Run regression method:
      SupportVectorSettings svmSettings = new SupportVectorSettings();
      svmSettings.setDataSpecification( msd.getMetaData() );

      // Assign settings:
      int iTar = msd.getMetaData().getAttributesNumber()-1;
      svmSettings.setTarget( msd.getMetaData().getMiningAttribute(iTar) );
      svmSettings.setSvmType( SupportVectorSettings.SVM_EPSILON_SVR );
      svmSettings.setKernelType( SupportVectorSettings.KERNEL_POLY);
      svmSettings.setC(1.0);
      svmSettings.setGamma(1.0);
      svmSettings.setDegree(2.0);
      svmSettings.setLossEpsilon(0.1);
      svmSettings.verifySettings();

      // Get default mining algorithm specification from 'algorithms.xml':
      MiningAlgorithmSpecification miningAlgorithmSpecification =
        MiningAlgorithmSpecification.getMiningAlgorithmSpecification( regressionMethod, null );
      if( miningAlgorithmSpecification == null )
        throw new MiningException( "Can't find application " + regressionMethod );

      // Get class name from algorithms specification:
      String className = miningAlgorithmSpecification.getClassname();
      if( className == null )
        throw new MiningException( "className attribute of regressionMethod expected." );

      // Create algorithm object with default values:
      MiningAlgorithm svmAlgo = GeneralUtils.createMiningAlgorithmInstance(className);

      // Put it all together:
      svmAlgo.setMiningInputStream( msd );
      svmAlgo.setMiningSettings( svmSettings );
      svmAlgo.setMiningAlgorithmSpecification( miningAlgorithmSpecification );
      svmAlgo.verify();

      // Build the mining model:
      msd.reset();
      svmm[i] = (SupportVectorMiningModel) svmAlgo.buildModel();
      RegressionDeviationAssessment rda = new RegressionDeviationAssessment();
      rda.setAssessmentData(msd);
      rda.setMiningModel(svmm[i]);
      msd.reset();
      double dev = rda.calculateAssessment();
      System.out.println("dev = " + dev);
    };
  }

  /**
   * Returns approximators for all attributes.
   *
   * @return array of approximators for all time series
   */
  protected SupervisedMiningModel[] getApproximator() {

    return svmm;
  }

  /**
   * Returns name of regression method for MiningAlgorithmSpecification.
   *
   * @return name of regression method
   */
  public String getRegressionMethod()
  {
    return regressionMethod;
  }

  /**
   * Sets new name of regression method for MiningAlgorithmSpecification.
   *
   * @param regressionMethod new name of regression method
   */
  public void setRegressionMethod(String regressionMethod)
  {
    this.regressionMethod = regressionMethod;
  }

}