linearnormal.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.Transform.OneToOne;

import com.prudsys.pdm.Adapters.PmmlVersion20.DerivedField;
import com.prudsys.pdm.Adapters.PmmlVersion20.LinearNorm;
import com.prudsys.pdm.Adapters.PmmlVersion20.NormContinuous;
import com.prudsys.pdm.Core.Category;
import com.prudsys.pdm.Core.MiningAttribute;
import com.prudsys.pdm.Core.MiningException;
import com.prudsys.pdm.Core.NumericAttribute;
import com.prudsys.pdm.Models.Statistics.Group;
import com.prudsys.pdm.Transform.OneToOneMapping;

/**
 * Realization of linear normalization. By default, the
 * (0,1) normalization is used.
 *
 * The value of minimum and maximum can be set via the
 * corresponding setter methods. If a statistics model
 * object is passed, they are taken from this.
 *
 * Missing values are transformed into missing values,
 * outliers can be handled on three different ways:
 * 1. as is, 2. as missing values, 3. as extreme values
 * with projection.
 */
public class LinearNormal extends OneToOneMapping
{
  // -----------------------------------------------------------------------
  //  Constants of types of outlier treatment
  // -----------------------------------------------------------------------
  /** Treatmeant of outliers as is. */
  public static final String OUTLIER_TREATMENT_METHOD_asIs            = "asIs";

  /** Treatment of outliers as missing values. */
  public static final String OUTLIER_TREATMENT_METHOD_asMissingValues = "asMissingValues";

  /** Treatment of values as extreme values, i.e. by projection to bounds. */
  public static final String OUTLIER_TREATMENT_METHOD_asExtremeValues = "asExtremeValues";

  /** Minimum span to be treated as positiv value. */
  public static final double IDENTITY_EPSILON = 0.00000000001;

  // -----------------------------------------------------------------------
  //  Variables declarations
  // -----------------------------------------------------------------------
  /** Minimum value of attribute. */
  private double min        = 0.0;

  /** Maximum value of attribute. */
  private double max        = 0.0;

  /** Lower bound of attribute. */
  private double lowerBound = 0.0;

  /** Upper bound of attribute. */
  private double upperBound = 1.0;

  /** Treatment of outliers, values outside interval [min, max]. */
  private String outliers   = OUTLIER_TREATMENT_METHOD_asIs;

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

  // -----------------------------------------------------------------------
  //  Getter and setter methods
  // -----------------------------------------------------------------------
  /**
   * Sets minimum value.
   *
   * @param min new minimum value
   */
  public void setMin(double min)
  {
    this.min = min;
  }

  /**
   * Returns minimum value.
   *
   * @return minimum value
   */
  public double getMin()
  {
    return min;
  }

  /**
   * Sets maximum value.
   *
   * @param max new maximum value
   */
  public void setMax(double max)
  {
    this.max = max;
  }

  /**
   * Returns maximum value.
   *
   * @return maximum value
   */
  public double getMax()
  {
    return max;
  }

  /**
   * Sets lower bound of normalized values (default 0).
   *
   * @param lowerBound new lower bound of normalized values
   */
  public void setLowerBound(double lowerBound)
  {
    this.lowerBound = lowerBound;
  }

  /**
   * Returns lower bound of normalized values (default 0).
   *
   * @return lower bound of normalized values
   */
  public double getLowerBound()
  {
    return lowerBound;
  }

  /**
   * Sets upper bound of normalized values (default 1).
   *
   * @param upperBound new upper bound of normalized values
   */
  public void setUpperBound(double upperBound)
  {
    this.upperBound = upperBound;
  }

  /**
   * Returns upper bound of normalized values (default 1).
   *
   * @return upper bound of normalized values
   */
  public double getUpperBound()
  {
    return upperBound;
  }

  /**
   * Sets treatment type of outliers.
   *
   * @param outliers new treatment type of outliers
   */
  public void setOutliers(String outliers)
  {
    this.outliers = outliers;
  }

  /**
   * Returns treatment type of outliers.
   *
   * @return treatment type of outliers
   */
  public String getOutliers()
  {
    return outliers;
  }

  // -----------------------------------------------------------------------
  //  Transformation methods
  // -----------------------------------------------------------------------
  /**
   * Transforms the source attribute. The result is the target attribute.
   *
   * @return transformed (normalized) attribute
   * @exception MiningException could not transform attribute
   */
  public MiningAttribute transformAttribute() throws MiningException
  {
      if (getSourceAttribute() == null)
        throw new MiningException("Could not find source attribute: " + sourceName);

      if (! (getSourceAttribute() instanceof NumericAttribute))
        throw new MiningException("Source attribute '" + sourceName + "' must be numeric");

      NumericAttribute sourceAttribute      = (NumericAttribute) getSourceAttribute();
      NumericAttribute transformedAttribute = (NumericAttribute) sourceAttribute.clone();

      transformedAttribute.setName( getTargetNameDynamic() );
      if (outliers == OUTLIER_TREATMENT_METHOD_asIs)
      {
      	  // For "As Is" the upper/lower bound is un-predictable, so set infinity
	      transformedAttribute.setLowerBound( Double.NEGATIVE_INFINITY );
	      transformedAttribute.setUpperBound( Double.POSITIVE_INFINITY );
      }else
      {
    	  transformedAttribute.setLowerBound( lowerBound );
	      transformedAttribute.setUpperBound( upperBound );
      }
      
      
      if (statisticsMiningModel != null) {
        Group group = statisticsMiningModel.getRootGroup();
        min = group.getMin();
        max = group.getMax();
      };

      return transformedAttribute;
  }

  /**
   * Transforms attribute value. The result is also a value.
   *
   * @param attributeValue value of attribute to be transformed
   * @return transformed (normalized) value
   * @exception MiningException could not transform attribute value
   */
  public double transformAttributeValue( double attributeValue ) throws MiningException
  {
      // Missing value:
      if (Category.isMissingValue(attributeValue))
        return attributeValue;

      // Outlier:
      if ( !outliers.equals(OUTLIER_TREATMENT_METHOD_asIs) ) {
        if (attributeValue < min || attributeValue > max) {
          if ( outliers.equals(OUTLIER_TREATMENT_METHOD_asMissingValues) )
            return Category.MISSING_VALUE;
          // Treat as extreme values:
          else {
            if (attributeValue < min)
              return lowerBound;   // projection to lower bound
            else
              return upperBound;   // projection to upper bound
          }
        }
      }

      // Normalization:
      if (Math.abs(max - min) < IDENTITY_EPSILON)
        return lowerBound;
      double transformedValue = lowerBound +
            (upperBound - lowerBound)*(attributeValue - min)/(max - min);

      return transformedValue;
  }

  // -----------------------------------------------------------------------
  //  Methods of PMML handling
  // -----------------------------------------------------------------------
  /**
   * Creates PMML object DerivedField of this object of NormContinuous type.
   *
   * @return DerivedField element
   * @see com.prudsys.pdm.Adapters.PmmlVersion20.NormContinuous
   * @exception MiningException could not create PMML object
   */
  public Object createPmmlObject() throws MiningException
  {
      DerivedField field = (DerivedField) super.createPmmlObject();

      NormContinuous normContinuous = new NormContinuous();
      normContinuous.setField( sourceName );
      LinearNorm linearNorm[] = new LinearNorm[2];
      linearNorm[0] = new LinearNorm();
      linearNorm[0].setOrig( "" + min );
      linearNorm[0].setNorm( String.valueOf(lowerBound) );
      linearNorm[1] = new LinearNorm();
      linearNorm[1].setOrig( "" + max );
      linearNorm[1].setNorm( String.valueOf(upperBound) );
      normContinuous.setLinearNorm( linearNorm );

      field.setNormContinuous( normContinuous );
      return field;
  }

  /**
   * Creates this object from PMML object DerivedField, subobject NormContinuous.
   *
   * @param pmml DerivedField element
   * @see com.prudsys.pdm.Adapters.PmmlVersion20.NormContinuous
   * @exception MiningException could not parse PMML object
   */
  public void parsePmmlObject(Object pmml) throws MiningException
  {
      super.parsePmmlObject(pmml);

      DerivedField field = (DerivedField) pmml;

      com.prudsys.pdm.Adapters.PmmlVersion20.NormContinuous norm = field.getNormContinuous();
      sourceName = norm.getField();
      LinearNorm linearNorm[] = norm.getLinearNorm();
      min        = Double.parseDouble( linearNorm[0].getOrig() );
      lowerBound = Double.parseDouble( linearNorm[0].getNorm() );
      max        = Double.parseDouble( linearNorm[1].getOrig() );
      upperBound = Double.parseDouble( linearNorm[1].getNorm() );
  }
}