principalcomponents.java

这是关于数据挖掘的一些算法

/*
 *    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.
 */

/*
 *    PrincipalComponents.java
 *    Copyright (C) 2000 University of Waikato, Hamilton, New Zealand
 *
 */

package weka.attributeSelection;

import weka.core.Attribute;
import weka.core.Capabilities;
import weka.core.FastVector;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.Matrix;
import weka.core.Option;
import weka.core.OptionHandler;
import weka.core.SparseInstance;
import weka.core.Utils;
import weka.core.Capabilities.Capability;
import weka.filters.Filter;
import weka.filters.unsupervised.attribute.NominalToBinary;
import weka.filters.unsupervised.attribute.Normalize;
import weka.filters.unsupervised.attribute.Remove;
import weka.filters.unsupervised.attribute.ReplaceMissingValues;

import java.util.Enumeration;
import java.util.Vector;

/**
 <!-- globalinfo-start -->
 * Performs a principal components analysis and transformation of the data. Use in conjunction with a Ranker search. Dimensionality reduction is accomplished by choosing enough eigenvectors to account for some percentage of the variance in the original data---default 0.95 (95%). Attribute noise can be filtered by transforming to the PC space, eliminating some of the worst eigenvectors, and then transforming back to the original space.
 * <p/>
 <!-- globalinfo-end -->
 *
 <!-- options-start -->
 * Valid options are: <p/>
 * 
 * <pre> -D
 *  Don't normalize input data.</pre>
 * 
 * <pre> -R
 *  Retain enough PC attributes to account 
 *  for this proportion of variance in the original data.
 *  (default = 0.95)</pre>
 * 
 * <pre> -O
 *  Transform through the PC space and 
 *  back to the original space.</pre>
 * 
 * <pre> -A
 *  Maximum number of attributes to include in 
 *  transformed attribute names. (-1 = include all)</pre>
 * 
 <!-- options-end -->
 *
 * @author Mark Hall ([e-mail])
 * @author Gabi Schmidberger ([e-mail])
 * @version $Revision: 1.35 $
 */
public class PrincipalComponents 
  extends UnsupervisedAttributeEvaluator 
  implements AttributeTransformer, OptionHandler {
  
  /** for serialization */
  static final long serialVersionUID = 3310137541055815078L;
  
  /** The data to transform analyse/transform */
  private Instances m_trainInstances;

  /** Keep a copy for the class attribute (if set) */
  private Instances m_trainCopy;

  /** The header for the transformed data format */
  private Instances m_transformedFormat;

  /** The header for data transformed back to the original space */
  private Instances m_originalSpaceFormat;

  /** Data has a class set */
  private boolean m_hasClass;

  /** Class index */
  private int m_classIndex;

  /** Number of attributes */
  private int m_numAttribs;

  /** Number of instances */
  private int m_numInstances;

  /** Correlation matrix for the original data */
  private double [][] m_correlation;

  /** Will hold the unordered linear transformations of the (normalized)
      original data */
  private double [][] m_eigenvectors;
  
  /** Eigenvalues for the corresponding eigenvectors */
  private double [] m_eigenvalues = null;

  /** Sorted eigenvalues */
  private int [] m_sortedEigens;

  /** sum of the eigenvalues */
  private double m_sumOfEigenValues = 0.0;
  
  /** Filters for original data */
  private ReplaceMissingValues m_replaceMissingFilter;
  private Normalize m_normalizeFilter;
  private NominalToBinary m_nominalToBinFilter;
  private Remove m_attributeFilter;
  
  /** used to remove the class column if a class column is set */
  private Remove m_attribFilter;

  /** The number of attributes in the pc transformed data */
  private int m_outputNumAtts = -1;
  
  /** normalize the input data? */
  private boolean m_normalize = true;

  /** the amount of varaince to cover in the original data when
      retaining the best n PC's */
  private double m_coverVariance = 0.95;

  /** transform the data through the pc space and back to the original
      space ? */
  private boolean m_transBackToOriginal = false;
  
  /** maximum number of attributes in the transformed attribute name */
  private int m_maxAttrsInName = 5;

  /** holds the transposed eigenvectors for converting back to the
      original space */
  private double [][] m_eTranspose;

  /**
   * Returns a string describing this attribute transformer
   * @return a description of the evaluator suitable for
   * displaying in the explorer/experimenter gui
   */
  public String globalInfo() {
    return "Performs a principal components analysis and transformation of "
      +"the data. Use in conjunction with a Ranker search. Dimensionality "
      +"reduction is accomplished by choosing enough eigenvectors to "
      +"account for some percentage of the variance in the original data---"
      +"default 0.95 (95%). Attribute noise can be filtered by transforming "
      +"to the PC space, eliminating some of the worst eigenvectors, and "
      +"then transforming back to the original space.";
  }

  /**
   * Returns an enumeration describing the available options. <p>
   *
   * @return an enumeration of all the available options.
   **/
  public Enumeration listOptions () {
    Vector newVector = new Vector(3);
    newVector.addElement(new Option("\tDon't normalize input data." 
                                    , "D", 0, "-D"));

    newVector.addElement(new Option("\tRetain enough PC attributes to account "
                                    +"\n\tfor this proportion of variance in "
                                    +"the original data.\n"
                                    + "\t(default = 0.95)",
                                    "R",1,"-R"));
    
    newVector.addElement(new Option("\tTransform through the PC space and "
                                    +"\n\tback to the original space."
                                    , "O", 0, "-O"));
                                    
    newVector.addElement(new Option("\tMaximum number of attributes to include in "
                                    + "\n\ttransformed attribute names. (-1 = include all)"
                                    , "A", 1, "-A"));
    return  newVector.elements();
  }

  /**
   * Parses a given list of options. <p/>
   *
   <!-- options-start -->
   * Valid options are: <p/>
   * 
   * <pre> -D
   *  Don't normalize input data.</pre>
   * 
   * <pre> -R
   *  Retain enough PC attributes to account 
   *  for this proportion of variance in the original data.
   *  (default = 0.95)</pre>
   * 
   * <pre> -O
   *  Transform through the PC space and 
   *  back to the original space.</pre>
   * 
   * <pre> -A
   *  Maximum number of attributes to include in 
   *  transformed attribute names. (-1 = include all)</pre>
   * 
   <!-- options-end -->
   *
   * @param options the list of options as an array of strings
   * @throws Exception if an option is not supported
   */
  public void setOptions (String[] options)
    throws Exception {
    resetOptions();
    String optionString;

    optionString = Utils.getOption('R', options);
    if (optionString.length() != 0) {
      Double temp;
      temp = Double.valueOf(optionString);
      setVarianceCovered(temp.doubleValue());
    }
    optionString = Utils.getOption('A', options);
    if (optionString.length() != 0) {
      setMaximumAttributeNames(Integer.parseInt(optionString));
    }
    setNormalize(!Utils.getFlag('D', options));

    setTransformBackToOriginal(Utils.getFlag('O', options));
  }

  /**
   * Reset to defaults
   */
  private void resetOptions() {
    m_coverVariance = 0.95;
    m_normalize = true;
    m_sumOfEigenValues = 0.0;
    m_transBackToOriginal = false;
  }

  /**
   * Returns the tip text for this property
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String normalizeTipText() {
    return "Normalize input data.";
  }

  /**
   * Set whether input data will be normalized.
   * @param n true if input data is to be normalized
   */
  public void setNormalize(boolean n) {
    m_normalize = n;
  }

  /**
   * Gets whether or not input data is to be normalized
   * @return true if input data is to be normalized
   */
  public boolean getNormalize() {
    return m_normalize;
  }

  /**
   * Returns the tip text for this property
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String varianceCoveredTipText() {
    return "Retain enough PC attributes to account for this proportion of "
      +"variance.";
  }

  /**
   * Sets the amount of variance to account for when retaining
   * principal components
   * @param vc the proportion of total variance to account for
   */
  public void setVarianceCovered(double vc) {
    m_coverVariance = vc;
  }

  /**
   * Gets the proportion of total variance to account for when
   * retaining principal components
   * @return the proportion of variance to account for
   */
  public double getVarianceCovered() {
    return m_coverVariance;
  }

  /**
   * Returns the tip text for this property
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String maximumAttributeNamesTipText() {
    return "The maximum number of attributes to include in transformed attribute names.";
  }

  /**
   * Sets maximum number of attributes to include in
   * transformed attribute names.
   * @param m the maximum number of attributes
   */
  public void setMaximumAttributeNames(int m) {
    m_maxAttrsInName = m;
  }

  /**
   * Gets maximum number of attributes to include in
   * transformed attribute names.
   * @return the maximum number of attributes
   */
  public int getMaximumAttributeNames() {
    return m_maxAttrsInName;
  }

  /**
   * Returns the tip text for this property
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String transformBackToOriginalTipText() {
    return "Transform through the PC space and back to the original space. "
      +"If only the best n PCs are retained (by setting varianceCovered < 1) "
      +"then this option will give a dataset in the original space but with "
      +"less attribute noise.";
  }

  /**
   * Sets whether the data should be transformed back to the original
   * space
   * @param b true if the data should be transformed back to the
   * original space
   */
  public void setTransformBackToOriginal(boolean b) {
    m_transBackToOriginal = b;
  }
  
  /**
   * Gets whether the data is to be transformed back to the original
   * space.
   * @return true if the data is to be transformed back to the original space
   */
  public boolean getTransformBackToOriginal() {
    return m_transBackToOriginal;
  }

  /**
   * Gets the current settings of PrincipalComponents
   *
   * @return an array of strings suitable for passing to setOptions()
   */
  public String[] getOptions () {

    String[] options = new String[6];
    int current = 0;

    if (!getNormalize()) {
      options[current++] = "-D";
    }

    options[current++] = "-R";
    options[current++] = ""+getVarianceCovered();

    options[current++] = "-A";
    options[current++] = ""+getMaximumAttributeNames();

    if (getTransformBackToOriginal()) {
      options[current++] = "-O";
    }
    
    while (current < options.length) {
      options[current++] = "";
    }
    
    return  options;
  }

  /**
   * Returns the capabilities of this evaluator.
   *
   * @return            the capabilities of this evaluator
   * @see               Capabilities
   */
  public Capabilities getCapabilities() {
    Capabilities result = super.getCapabilities();
    
    // attributes
    result.enable(Capability.NOMINAL_ATTRIBUTES);
    result.enable(Capability.NUMERIC_ATTRIBUTES);
    result.enable(Capability.DATE_ATTRIBUTES);
    result.enable(Capability.MISSING_VALUES);
    
    // class
    result.enable(Capability.NOMINAL_CLASS);
    result.enable(Capability.NUMERIC_CLASS);
    result.enable(Capability.DATE_CLASS);
    result.enable(Capability.MISSING_CLASS_VALUES);
    result.enable(Capability.NO_CLASS);
    
    return result;
  }

  /**
   * Initializes principal components and performs the analysis
   * @param data the instances to analyse/transform
   * @throws Exception if analysis fails
   */
  public void buildEvaluator(Instances data) throws Exception {
    // can evaluator handle data?
    getCapabilities().testWithFail(data);

    buildAttributeConstructor(data);
  }

  private void buildAttributeConstructor (Instances data) throws Exception {
    m_eigenvalues = null;
    m_outputNumAtts = -1;
    m_attributeFilter = null;
    m_nominalToBinFilter = null;
    m_sumOfEigenValues = 0.0;
    m_trainInstances = new Instances(data);

    // make a copy of the training data so that we can get the class
    // column to append to the transformed data (if necessary)
    m_trainCopy = new Instances(m_trainInstances);
    
    m_replaceMissingFilter = new ReplaceMissingValues();
    m_replaceMissingFilter.setInputFormat(m_trainInstances);
    m_trainInstances = Filter.useFilter(m_trainInstances, 
                                        m_replaceMissingFilter);

    if (m_normalize) {
      m_normalizeFilter = new Normalize();
      m_normalizeFilter.setInputFormat(m_trainInstances);
      m_trainInstances = Filter.useFilter(m_trainInstances, m_normalizeFilter);
    }

    m_nominalToBinFilter = new NominalToBinary();
    m_nominalToBinFilter.setInputFormat(m_trainInstances);
    m_trainInstances = Filter.useFilter(m_trainInstances, 
                                        m_nominalToBinFilter);
    
    // delete any attributes with only one distinct value or are all missing
    Vector deleteCols = new Vector();
    for (int i=0;i<m_trainInstances.numAttributes();i++) {
      if (m_trainInstances.numDistinctValues(i) <=1) {
        deleteCols.addElement(new Integer(i));
      }
    }

    if (m_trainInstances.classIndex() >=0) {
      // get rid of the class column
      m_hasClass = true;
      m_classIndex = m_trainInstances.classIndex();
      deleteCols.addElement(new Integer(m_classIndex));
    }

    // remove columns from the data if necessary
    if (deleteCols.size() > 0) {
      m_attributeFilter = new Remove();
      int [] todelete = new int [deleteCols.size()];