gridsearch.java

代码是一个分类器的实现,其中使用了部分weka的源代码。可以将项目导入eclipse运行

/*
 *    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.
 */

/*
 * GridSearch.java
 * Copyright (C) 2006 University of Waikato, Hamilton, New Zealand
 */

package weka.classifiers.meta;

import weka.classifiers.Classifier;
import weka.classifiers.Evaluation;
import weka.classifiers.RandomizableSingleClassifierEnhancer;
import weka.classifiers.functions.LinearRegression;
import weka.core.AdditionalMeasureProducer;
import weka.core.Capabilities;
import weka.core.Debug;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.MathematicalExpression;
import weka.core.Option;
import weka.core.OptionHandler;
import weka.core.PropertyPath;
import weka.core.SelectedTag;
import weka.core.SerializedObject;
import weka.core.Tag;
import weka.core.Utils;
import weka.core.Capabilities.Capability;
import weka.filters.Filter;
import weka.filters.supervised.attribute.PLSFilter;
import weka.filters.unsupervised.attribute.MathExpression;
import weka.filters.unsupervised.attribute.NumericCleaner;
import weka.filters.unsupervised.instance.Resample;

import java.beans.PropertyDescriptor;
import java.io.File;
import java.io.Serializable;
import java.util.Collections;
import java.util.Comparator;
import java.util.Enumeration;
import java.util.HashMap;
import java.util.Hashtable;
import java.util.Iterator;
import java.util.Random;
import java.util.Vector;

/**
 <!-- globalinfo-start -->
 * Performs a grid search of parameter pairs for the a classifier (Y-axis, default is LinearRegression with the "Ridge" parameter) and the PLSFilter (X-axis, "# of Components") and chooses the best pair found for the actual predicting.<br/>
 * <br/>
 * The initial grid is worked on with 2-fold CV to determine the values of the parameter pairs for the selected type of evaluation (e.g., accuracy). The best point in the grid is then taken and a 10-fold CV is performed with the adjacent parameter pairs. If a better pair is found, then this will act as new center and another 10-fold CV will be performed (kind of hill-climbing). This process is repeated until no better pair is found or the best pair is on the border of the grid.<br/>
 * In case the best pair is on the border, one can let GridSearch automatically extend the grid and continue the search. Check out the properties 'gridIsExtendable' (option '-extend-grid') and 'maxGridExtensions' (option '-max-grid-extensions &lt;num&gt;').<br/>
 * <br/>
 * GridSearch can handle doubles, integers (values are just cast to int) and booleans (0 is false, otherwise true). float, char and long are supported as well.<br/>
 * <br/>
 * The best filter/classifier setup can be accessed after the buildClassifier call via the getBestFilter/getBestClassifier methods.<br/>
 * Note on the implementation: after the data has been passed through the filter, a default NumericCleaner filter is applied to the data in order to avoid numbers that are getting too small and might produce NaNs in other schemes.
 * <p/>
 <!-- globalinfo-end -->
 * 
 <!-- options-start -->
 * Valid options are: <p/>
 * 
 * <pre> -E &lt;CC|RMSE|RRSE|MAE|RAE|COMB|ACC&gt;
 *  Determines the parameter used for evaluation:
 *  CC = Correlation coefficient
 *  RMSE = Root mean squared error
 *  RRSE = Root relative squared error
 *  MAE = Mean absolute error
 *  RAE = Root absolute error
 *  COMB = Combined = (1-abs(CC)) + RRSE + RAE
 *  ACC = Accuracy
 *  (default: CC)</pre>
 * 
 * <pre> -y-property &lt;option&gt;
 *  The Y option to test (without leading dash).
 *  (default: classifier.ridge)</pre>
 * 
 * <pre> -y-min &lt;num&gt;
 *  The minimum for Y.
 *  (default: -10)</pre>
 * 
 * <pre> -y-max &lt;num&gt;
 *  The maximum for Y.
 *  (default: +5)</pre>
 * 
 * <pre> -y-step &lt;num&gt;
 *  The step size for Y.
 *  (default: 1)</pre>
 * 
 * <pre> -y-base &lt;num&gt;
 *  The base for Y.
 *  (default: 10)</pre>
 * 
 * <pre> -y-expression &lt;expr&gt;
 *  The expression for Y.
 *  Available parameters:
 *   BASE
 *   FROM
 *   TO
 *   STEP
 *   I - the current iteration value
 *   (from 'FROM' to 'TO' with stepsize 'STEP')
 *  (default: 'pow(BASE,I)')</pre>
 * 
 * <pre> -filter &lt;filter specification&gt;
 *  The filter to use (on X axis). Full classname of filter to include, 
 *  followed by scheme options.
 *  (default: weka.filters.supervised.attribute.PLSFilter)</pre>
 * 
 * <pre> -x-property &lt;option&gt;
 *  The X option to test (without leading dash).
 *  (default: filter.numComponents)</pre>
 * 
 * <pre> -x-min &lt;num&gt;
 *  The minimum for X.
 *  (default: +5)</pre>
 * 
 * <pre> -x-max &lt;num&gt;
 *  The maximum for X.
 *  (default: +20)</pre>
 * 
 * <pre> -x-step &lt;num&gt;
 *  The step size for X.
 *  (default: 1)</pre>
 * 
 * <pre> -x-base &lt;num&gt;
 *  The base for X.
 *  (default: 10)</pre>
 * 
 * <pre> -x-expression &lt;expr&gt;
 *  The expression for the X value.
 *  Available parameters:
 *   BASE
 *   MIN
 *   MAX
 *   STEP
 *   I - the current iteration value
 *   (from 'FROM' to 'TO' with stepsize 'STEP')
 *  (default: 'pow(BASE,I)')</pre>
 * 
 * <pre> -extend-grid
 *  Whether the grid can be extended.
 *  (default: no)</pre>
 * 
 * <pre> -max-grid-extensions &lt;num&gt;
 *  The maximum number of grid extensions (-1 is unlimited).
 *  (default: 3)</pre>
 * 
 * <pre> -sample-size &lt;num&gt;
 *  The size (in percent) of the sample to search the inital grid with.
 *  (default: 100)</pre>
 * 
 * <pre> -traversal &lt;ROW-WISE|COLUMN-WISE&gt;
 *  The type of traversal for the grid.
 *  (default: COLUMN-WISE)</pre>
 * 
 * <pre> -log-file &lt;filename&gt;
 *  The log file to log the messages to.
 *  (default: none)</pre>
 * 
 * <pre> -S &lt;num&gt;
 *  Random number seed.
 *  (default 1)</pre>
 * 
 * <pre> -D
 *  If set, classifier is run in debug mode and
 *  may output additional info to the console</pre>
 * 
 * <pre> -W
 *  Full name of base classifier.
 *  (default: weka.classifiers.functions.LinearRegression)</pre>
 * 
 * <pre> 
 * Options specific to classifier weka.classifiers.functions.LinearRegression:
 * </pre>
 * 
 * <pre> -D
 *  Produce debugging output.
 *  (default no debugging output)</pre>
 * 
 * <pre> -S &lt;number of selection method&gt;
 *  Set the attribute selection method to use. 1 = None, 2 = Greedy.
 *  (default 0 = M5' method)</pre>
 * 
 * <pre> -C
 *  Do not try to eliminate colinear attributes.
 * </pre>
 * 
 * <pre> -R &lt;double&gt;
 *  Set ridge parameter (default 1.0e-8).
 * </pre>
 * 
 * <pre> 
 * Options specific to filter weka.filters.supervised.attribute.PLSFilter ('-filter'):
 * </pre>
 * 
 * <pre> -D
 *  Turns on output of debugging information.</pre>
 * 
 * <pre> -C &lt;num&gt;
 *  The number of components to compute.
 *  (default: 20)</pre>
 * 
 * <pre> -U
 *  Updates the class attribute as well.
 *  (default: off)</pre>
 * 
 * <pre> -M
 *  Turns replacing of missing values on.
 *  (default: off)</pre>
 * 
 * <pre> -A &lt;SIMPLS|PLS1&gt;
 *  The algorithm to use.
 *  (default: PLS1)</pre>
 * 
 * <pre> -P &lt;none|center|standardize&gt;
 *  The type of preprocessing that is applied to the data.
 *  (default: center)</pre>
 * 
 <!-- options-end -->
 *
 * Examples:
 * <ul>
 *   <li>
 *     <b>Optimizing SMO with RBFKernel (C and gamma)</b>
 *     <ul>
 *       <li>Set the evaluation to <i>Accuracy</i>.</li>
 *       <li>Set the filter to <code>weka.filters.AllFilter</code> since we
 *           don't need any special data processing and we don't optimize the
 *           filter in this case (data gets always passed through filter!).</li>
 *       <li>Set <code>weka.classifiers.functions.SMO</code> as classifier
 *           with <code>weka.classifiers.functions.supportVector.RBFKernel</code>
 *           as kernel.
 *       </li>
 *       <li>Set the XProperty to "classifier.c", XMin to "1", XMax to "16", 
 *           XStep to "1" and the XExpression to "I". This will test the "C"
 *           parameter of SMO for the values from 1 to 16.</li>
 *       <li>Set the YProperty to "classifier.kernel.gamma", YMin to "-5",
 *           YMax to "2", YStep to "1" YBase to "10" and YExpression to 
 *           "pow(BASE,I)". This will test the gamma of the RBFKernel with the
 *           values 10^-5, 10^-4,..,10^2.</li>
 *     </ul>
 *   </li>
 *   <li>
 *     <b>Optimizing PLSFilter with LinearRegression (# of components and ridge) - default setup</b>
 *     <ul>
 *       <li>Set the evaluation to <i>Correlation coefficient</i>.</li>
 *       <li>Set the filter to <code>weka.filters.supervised.attribute.PLSFilter</code>.</li>
 *       <li>Set <code>weka.classifiers.functions.LinearRegression</code> as 
 *           classifier and use no attribute selection and no elimination of
 *           colinear attributes.</li>
 *       <li>Set the XProperty to "filter.numComponents", XMin to "5", XMax 
 *           to "20" (this depends heavily on your dataset, should be no more
 *           than the number of attributes!), XStep to "1" and XExpression to
 *           "I". This will test the number of components the PLSFilter will
 *           produce from 5 to 20.</li>
 *       <li>Set the YProperty to "classifier.ridge", XMin to "-10", XMax to 
 *           "5", YStep to "1" and YExpression to "pow(BASE,I)". This will
 *           try ridge parameters from 10^-10 to 10^5.</li>
 *     </ul>
 *   </li>
 * </ul>
 * 
 * General notes:
 * <ul>
 *   <li>Turn the <i>debug</i> flag on in order to see some progress output in the
 *       console</li>
 *   <li>If you want to view the fitness landscape that GridSearch explores,
 *       select a <i>log file</i>. This log will then contain Gnuplot data and 
 *       script block for viewing the landscape. Just copy paste those blocks 
 *       into files named accordingly and run Gnuplot with them.</li>
 * </ul>
 *
 * @author  Bernhard Pfahringer (bernhard at cs dot waikato dot ac dot nz)
 * @author  Geoff Holmes (geoff at cs dot waikato dot ac dot nz)
 * @author  fracpete (fracpete at waikato dot ac dot nz)
 * @version $Revision: 1.3 $
 * @see     PLSFilter
 * @see     LinearRegression
 * @see	    NumericCleaner
 */
public class GridSearch
  extends RandomizableSingleClassifierEnhancer
  implements AdditionalMeasureProducer {

  /**
   * a serializable version of Point2D.Double
   * 
   * @see java.awt.geom.Point2D.Double
   */
  protected class PointDouble
    extends java.awt.geom.Point2D.Double
    implements Serializable {

    /** for serialization */
    private static final long serialVersionUID = 7151661776161898119L;
    
    /**
     * the default constructor
     * 
     * @param x		the x value of the point
     * @param y		the y value of the point
     */
    public PointDouble(double x, double y) {
      super(x, y);
    }
    
    /**
     * returns a string representation of the Point
     * 
     * @return the point as string
     */
    public String toString() {
      return super.toString().replaceAll(".*\\[", "[");
    }
  }

  /**
   * a serializable version of Point
   * 
   * @see java.awt.Point
   */
  protected class PointInt
    extends java.awt.Point
    implements Serializable {

    /** for serialization */
    private static final long serialVersionUID = -5900415163698021618L;

    /**
     * the default constructor
     * 
     * @param x		the x value of the point
     * @param y		the y value of the point
     */
    public PointInt(int x, int y) {
      super(x, y);
    }
    
    /**
     * returns a string representation of the Point
     * 
     * @return the point as string
     */
    public String toString() {
      return super.toString().replaceAll(".*\\[", "[");
    }
  }
  
  /**
   * for generating the parameter pairs in a grid
   */
  protected class Grid
    implements Serializable {

    /** for serialization */
    private static final long serialVersionUID = 7290732613611243139L;
    
    /** the minimum on the X axis */
    protected double m_MinX;
    
    /** the maximum on the X axis */
    protected double m_MaxX;
    
    /** the step size for the X axis */
    protected double m_StepX;

    /** the label for the X axis */
    protected String m_LabelX;
    
    /** the minimum on the Y axis */
    protected double m_MinY;
    
    /** the maximum on the Y axis */
    protected double m_MaxY;
    
    /** the step size for the Y axis */
    protected double m_StepY;

    /** the label for the Y axis */
    protected String m_LabelY;
    
    /** the number of points on the X axis */
    protected int m_Width;
    
    /** the number of points on the Y axis */
    protected int m_Height;
    
    /**
     * initializes the grid
     * 
     * @param minX 	the minimum on the X axis
     * @param maxX 	the maximum on the X axis
     * @param stepX 	the step size for the X axis
     * @param minY 	the minimum on the Y axis
     * @param maxY 	the maximum on the Y axis
     * @param stepY 	the step size for the Y axis
     */
    public Grid(double minX, double maxX, double stepX, 
	        double minY, double maxY, double stepY) {
      this(minX, maxX, stepX, "", minY, maxY, stepY, "");
    }

    
    /**
     * initializes the grid
     * 
     * @param minX 	the minimum on the X axis
     * @param maxX 	the maximum on the X axis
     * @param stepX 	the step size for the X axis
     * @param labelX	the label for the X axis
     * @param minY 	the minimum on the Y axis
     * @param maxY 	the maximum on the Y axis
     * @param stepY 	the step size for the Y axis
     * @param labelY	the label for the Y axis
     */
    public Grid(double minX, double maxX, double stepX, String labelX,
	        double minY, double maxY, double stepY, String labelY) {

      super();
      
      m_MinX   = minX;
      m_MaxX   = maxX;
      m_StepX  = stepX;
      m_LabelX = labelX;
      m_MinY   = minY;
      m_MaxY   = maxY;
      m_StepY  = stepY;
      m_LabelY = labelY;
      m_Height = (int) StrictMath.round((m_MaxY - m_MinY) / m_StepY) + 1;
      m_Width  = (int) StrictMath.round((m_MaxX - m_MinX) / m_StepX) + 1;
    }
    
    /**
     * returns the left border
     * 
     * @return 		the left border
     */
    public double getMinX() {
      return m_MinX;
    }
    
    /**
     * returns the right border
     *