treatoutlierattributevalue.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.Core.CategoricalAttribute;
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.Transform.OneToOneMapping;

/**
 * Treatment of outliers. By default, the
 * 'asIs' treatment is used which does nothing.
 *
 * For identifying whether the value is an outlier,
 * an assessment attribute must be defined which is
 * of the same type like the attribute to be transfomed.
 * For numeric attributes the value is tested for the
 * interval (lowerBound, upperBound), for categorical
 * attributes it must correspond to one of its categories.
 *
 * Use identical attribute transformation; hence target
 * name of attribute is identical to source name and
 * therefore ignored.
 *
 * No PMML convertation implemented because this is usually done
 * via the PMML element MiningSchema.
 */
public class TreatOutlierAttributeValue extends OneToOneMapping
{
  // -----------------------------------------------------------------------
  //  Constants of types of outliers trealment
  // -----------------------------------------------------------------------
  public static final String OUTLIER_TREATMENT_METHOD_asIs            = "asIs";
  public static final String OUTLIER_TREATMENT_METHOD_asMissingValues = "asMissingValues";
  public static final String OUTLIER_TREATMENT_METHOD_asExtremeValues = "asExtremeValues";

  // -----------------------------------------------------------------------
  //  Variables declarations
  // -----------------------------------------------------------------------
  /** Treatment type of outliers. */
  private String outliers = OUTLIER_TREATMENT_METHOD_asIs;

  /** Replacement value for lower projection for 'asExtremeValues'. */
  private double lowValue = 0.0;

  /** Replacement value for upper projection for 'asExtremeValues'. */
  private double highValue = 0.0;

  /** Attribute for evaluating if value is an outlier. */
  private MiningAttribute assessmentAttribute = null;

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

  // -----------------------------------------------------------------------
  //  Getter and setter methods
  // -----------------------------------------------------------------------
  /**
   * Returns type of outlier treatment.
   *
   * @return type of outlier treatment
   */
  public String getOutliers()
  {
    return outliers;
  }

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

  /**
   * Returns replacement value for lower projection for type 'asExtremeValues'.
   *
   * @return lower replacement value
   */
  public double getLowValue()
  {
    return lowValue;
  }

  /**
   * Sets replacement value for lower projection for type 'asExtremeValues'.
   *
   * @param lowValue new low replacement value
   */
  public void setLowValue(double lowValue)
  {
    this.lowValue = lowValue;
  }

  /**
   * Returns replacement value for upper projection for type 'asExtremeValues'.
   *
   * @return upper replacement value
   */
  public double getHighValue()
  {
    return highValue;
  }

  /**
   * Sets replacement value for upper projection for type 'asExtremeValues'.
   *
   * @param highValue new upper replacement value
   */
  public void setHighValue(double highValue)
  {
    this.highValue = highValue;
  }

  /**
   * Returns attribute used for outlier testing.
   *
   * @return attribute used for outlier test
   */
  public MiningAttribute getAssessmentAttribute()
  {
    return assessmentAttribute;
  }

  /**
   * Sets attribute used for outlier test.
   *
   * @param assessmentAttribute new attribute used for outlier test
   */
  public void setAssessmentAttribute(MiningAttribute assessmentAttribute)
  {
    this.assessmentAttribute = assessmentAttribute;
  }

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

    try
	{
    	MiningAttribute att = (MiningAttribute)sourceAttribute.clone();
        return att;
	}catch (Exception e)
	{
		throw new MiningException(e.getMessage());
	}
  }

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

      // Treatment 'asIs' => return:
      if (outliers.equals(OUTLIER_TREATMENT_METHOD_asIs))
        return transformedValue;

      // Is missing value or no assessment attribute => return:
      if (Category.isMissingValue( attributeValue ) || assessmentAttribute == null)
        return transformedValue;

      // Test for outlier:
      int outlierType = 0;
      MiningAttribute att = getSourceAttribute();
      if (att instanceof CategoricalAttribute) {
        if (assessmentAttribute instanceof NumericAttribute)
          throw new MiningException("Assessment attribute '" + assessmentAttribute + "' must be categorical");
        Category categ = ((CategoricalAttribute) att).getCategory(attributeValue);
        double key     = ((CategoricalAttribute) assessmentAttribute).getKey(categ);
        if (Category.isMissingValue(key))
          outlierType = 3;
      }
      else {
        if (assessmentAttribute instanceof CategoricalAttribute)
          throw new MiningException("Assessment attribute '" + assessmentAttribute + "' must be numeric");
        NumericAttribute numAtt = (NumericAttribute) assessmentAttribute;
        if (attributeValue < numAtt.getLowerBound())
          outlierType = 1;
        if (attributeValue > numAtt.getUpperBound())
          outlierType = 2;
      };

      // Treatment of outlier:
      if (outlierType > 0) {
        if ( outliers.equals(OUTLIER_TREATMENT_METHOD_asMissingValues) )
          transformedValue = Category.MISSING_VALUE;
        else {
          if (outlierType == 1) transformedValue = lowValue;
          if (outlierType == 2) transformedValue = highValue;
          if (outlierType == 3) ;  // meaningless treatment definition
        }
      };

      return transformedValue;
  }

  // -----------------------------------------------------------------------
  //  Methods of PMML handling
  // -----------------------------------------------------------------------
  /**
   * Creates PMML object DerivedField of this object.
   *
   * @return DerivedField element
   * @see com.prudsys.pdm.Adapters.PmmlVersion20.TransformationDictionary
   * @exception MiningException could not create PMML object
   */
  public Object createPmmlObject() throws MiningException
  {
    /**@todo Implement this com.prudsys.pdm.DataMining.PmmlPresentable method*/
      throw new java.lang.UnsupportedOperationException("Method createPmmlObject() not yet implemented.");
  }

  /**
   * Creates this object from PMML.
   * Currently not supported.
   *
   * @param pmml pmml element
   * @exception MiningException always thrown
   */
  public void parsePmmlObject(Object pmml) throws MiningException
  {
      /**@todo Implement this com.prudsys.pdm.DataMining.PmmlPresentable method*/
      throw new java.lang.UnsupportedOperationException("Method parsePmmlObject() not yet implemented.");
  }
}