timeseriesminingalgorithm.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 1.0
 */

package com.prudsys.pdm.Models.TimeSeriesPredict;

import com.prudsys.pdm.Core.ApplicationAttribute;
import com.prudsys.pdm.Core.AttributeType;
import com.prudsys.pdm.Core.CategoricalAttribute;
import com.prudsys.pdm.Core.MiningAlgorithm;
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.Supervised.SupervisedMiningModel;
import com.prudsys.pdm.Transform.Special.ReplaceMissingValueStream;

/**
 * Base class for time series mining algorithms.
 */
public abstract class TimeSeriesMiningAlgorithm extends MiningAlgorithm
{
    protected int embeddingDimension = 5;

    protected int stepSize = 1;

    protected boolean singleApproximator = true;

    /**
     * Empty constructor.
     */
    public TimeSeriesMiningAlgorithm()
    {
    }

    /**
     * Sets time series mining settings.
     *
     * @param miningSettings new time series mining settings
     * @exception IllegalArgumentException mining settings not time series mining settings
     */
    public void setMiningSettings( MiningSettings miningSettings ) throws IllegalArgumentException
    {
        if( miningSettings instanceof TimeSeriesMiningSettings )
        {
            super.setMiningSettings( miningSettings );
            TimeSeriesMiningSettings timeSeriesMiningSettings = (TimeSeriesMiningSettings)miningSettings;
            this.embeddingDimension = timeSeriesMiningSettings.getEmbeddingDimension();
            this.stepSize = timeSeriesMiningSettings.getStepSize();
            this.singleApproximator = timeSeriesMiningSettings.isSingleApproximator();
        }
        else
        {
            throw new IllegalArgumentException( "MiningSettings have to be TimeSeriesMiningSettings." );
        }
    }

    /**
     * Builds mining model by running the time series algorithm internally.
     * Before starting the algorithm, missing values are replaced.
     *
     * @return time series mining model generated by the algorithm
     * @exception MiningException error while running algorithm
     */
    public MiningModel buildModel() throws MiningException {

      long start = ( new java.util.Date() ).getTime();

      // Replace missing values by mean and mode values:
      ReplaceMissingValueStream rep = new ReplaceMissingValueStream(miningInputStream);
      miningInputStream             = new MiningArrayStream( rep.createReplaceMissingValueStream() );

      // Run time series algorithm:
      runAlgorithm();

      // Build time series model:
      TimeSeriesMiningModel model = new TimeSeriesMiningModel();
      model.setMiningSettings( miningSettings );
      model.setInputSpec( applicationInputSpecification );
      model.setMiningTransform( rep.getMts() );

      // Missing values in application input specification:
      ApplicationAttribute[] appAtt = applicationInputSpecification.getInputAttribute();
      double[] repVal = rep.getRepValues();
      for (int i = 0; i < appAtt.length; i++) {
        if (appAtt[i].getAttributeType().getType() == AttributeType.NUMERICAL) {
          appAtt[i].setMissingValueTreatment(
            ApplicationAttribute.MISSING_VALUE_TREATMENT_METHOD_asMean);
          appAtt[i].setMissingValueReplacement( String.valueOf(repVal[i]) );
        };
        if (appAtt[i].getAttributeType().getType() == AttributeType.CATEGORICAL) {
          appAtt[i].setMissingValueTreatment(
            ApplicationAttribute.MISSING_VALUE_TREATMENT_METHOD_asMode);
          appAtt[i].setMissingValueReplacement(
            ((CategoricalAttribute) metaData.getMiningAttribute(i)).getCategory( repVal[i] ).getDisplayValue() );
        };
      };

      // Set time series parameter:
      model.setEmbeddingDimension( embeddingDimension );
      model.setStepSize( stepSize );
      model.setSingleApproximator( singleApproximator );

      // Set approximators:
      model.setApproximator( getApproximator() );

      this.miningModel = model;

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

      return model;
    }

    /**
     * Runs time series mining algorithm.
     *
     * @exception MiningException error while running algorithm
     */
    protected abstract void runAlgorithm() throws MiningException;

    /**
     * Returns approximators for all attributes.
     *
     * @return approximators of allattributes
     */
    protected abstract SupervisedMiningModel[] getApproximator();

    /**
     * Creates an instance of the time series mining settings class that is
     * required to run the algorithm. The mining settings are assigned through
     * the setMiningSettings method.
     *
     * @return new instance of the time series mining settings class of the algorithm
     */
    public MiningSettings createMiningSettings() {

      return new TimeSeriesMiningSettings();
    }

}