weka.attributeSelection
Class RandomSearch

java.lang.Object
  weka.attributeSelection.ASSearch
      weka.attributeSelection.RandomSearch

All Implemented Interfaces:

OptionHandler, java.io.Serializable, StartSetHandler

public class RandomSearch

extends ASSearch

implements StartSetHandler, OptionHandler

Class for performing a random search.

Valid options are:

-P
Specify a starting set of attributes. Eg 1,4,7-9.

-F Percentage of the search space to consider. (default = 25).

-V
Verbose output. Output new best subsets as the search progresses.

Version:

$Revision: 1.12 $

Author:

Mark Hall (mhall@cs.waikato.ac.nz)

See Also:

Serialized Form

Field Summary

private java.util.BitSet	m_bestGroup the best feature set found during the search
private double	m_bestMerit the merit of the best subset found
private int	m_classIndex holds the class index
private boolean	m_hasClass does the data have a class
private int	m_iterations the number of iterations performed
private int	m_numAttribs number of attributes in the data
private boolean	m_onlyConsiderBetterAndSmaller only accept a feature set as being "better" than the best if its merit is better or equal to the best, and it contains fewer features than the best (this allows LVF to be implimented).
private java.util.Random	m_random random number object
private double	m_searchSize percentage of the search space to consider
private int	m_seed seed for random number generation
private int[]	m_starting holds a starting set as an array of attributes.
private Range	m_startRange holds the start set as a range
private boolean	m_verbose output new best subsets as the search progresses

Constructor Summary

RandomSearch()
Constructor

Method Summary

private int[]	attributeList(java.util.BitSet group) converts a BitSet into a list of attribute indexes
private int	countFeatures(java.util.BitSet featureSet) counts the number of features in a subset
private java.util.BitSet	generateRandomSubset() generates a random subset
java.lang.String[]	getOptions() Gets the current settings of RandomSearch.
double	getSearchPercent() get the percentage of the search space to consider
java.lang.String	getStartSet() Returns a list of attributes (and or attribute ranges) as a String
boolean	getVerbose() get whether or not output is verbose
java.lang.String	globalInfo() Returns a string describing this search method
java.util.Enumeration	listOptions() Returns an enumeration describing the available options.
private java.lang.String	printSubset(java.util.BitSet temp) prints a subset as a series of attribute numbers
private void	resetOptions() resets to defaults
int[]	search(ASEvaluation ASEval, Instances data) Searches the attribute subset space randomly.
java.lang.String	searchPercentTipText() Returns the tip text for this property
void	setOptions(java.lang.String[] options) Parses a given list of options.
void	setSearchPercent(double p) set the percentage of the search space to consider
void	setStartSet(java.lang.String startSet) Sets a starting set of attributes for the search.
void	setVerbose(boolean v) set whether or not to output new best subsets as the search proceeds
java.lang.String	startSetTipText() Returns the tip text for this property
private java.lang.String	startSetToString() converts the array of starting attributes to a string.
java.lang.String	toString() prints a description of the search
java.lang.String	verboseTipText() Returns the tip text for this property

Methods inherited from class weka.attributeSelection.ASSearch

forName

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Detail

m_starting

private int[] m_starting

holds a starting set as an array of attributes.

m_startRange

private Range m_startRange

holds the start set as a range

m_bestGroup

private java.util.BitSet m_bestGroup

the best feature set found during the search

m_bestMerit

private double m_bestMerit

the merit of the best subset found

m_onlyConsiderBetterAndSmaller

private boolean m_onlyConsiderBetterAndSmaller

only accept a feature set as being "better" than the best if its merit is better or equal to the best, and it contains fewer features than the best (this allows LVF to be implimented).

m_hasClass

private boolean m_hasClass

does the data have a class

m_classIndex

private int m_classIndex

holds the class index

m_numAttribs

private int m_numAttribs

number of attributes in the data

m_seed

private int m_seed

seed for random number generation

m_searchSize

private double m_searchSize

percentage of the search space to consider

m_iterations

private int m_iterations

the number of iterations performed

m_random

private java.util.Random m_random

random number object

m_verbose

private boolean m_verbose

output new best subsets as the search progresses

Constructor Detail

RandomSearch

public RandomSearch()

Constructor

Method Detail

globalInfo

public java.lang.String globalInfo()

Returns a string describing this search method

Returns:

a description of the search suitable for displaying in the explorer/experimenter gui

listOptions

public java.util.Enumeration listOptions()

Returns an enumeration describing the available options.

Specified by:

listOptions in interface OptionHandler

Returns:

an enumeration of all the available options.

setOptions

public void setOptions(java.lang.String[] options)
                throws java.lang.Exception

Parses a given list of options. Valid options are:

-P
Specify a starting set of attributes. Eg 1,4,7-9.

-F Percentage of the search space to consider. (default = 25).

-V
Verbose output. Output new best subsets as the search progresses.

Specified by:

setOptions in interface OptionHandler

Parameters:

options - the list of options as an array of strings

Throws:

java.lang.Exception - if an option is not supported

startSetTipText

public java.lang.String startSetTipText()

Returns the tip text for this property

Returns:

tip text for this property suitable for displaying in the explorer/experimenter gui

setStartSet

public void setStartSet(java.lang.String startSet)
                 throws java.lang.Exception

Sets a starting set of attributes for the search. It is the search method's responsibility to report this start set (if any) in its toString() method.

Specified by:

setStartSet in interface StartSetHandler

Parameters:

startSet - a string containing a list of attributes (and or ranges), eg. 1,2,6,10-15. "" indicates no start point. If a start point is supplied, random search evaluates the start point and then looks for subsets that are as good as or better than the start point with the same or lower cardinality.

Throws:

java.lang.Exception - if start set can't be set.

getStartSet

public java.lang.String getStartSet()

Returns a list of attributes (and or attribute ranges) as a String

Specified by:

getStartSet in interface StartSetHandler

Returns:

a list of attributes (and or attribute ranges)

verboseTipText

public java.lang.String verboseTipText()

Returns the tip text for this property

Returns:

tip text for this property suitable for displaying in the explorer/experimenter gui

setVerbose

public void setVerbose(boolean v)

set whether or not to output new best subsets as the search proceeds

Parameters:

v - true if output is to be verbose

getVerbose

public boolean getVerbose()

get whether or not output is verbose

Returns:

true if output is set to verbose

searchPercentTipText

public java.lang.String searchPercentTipText()

Returns the tip text for this property

Returns:

tip text for this property suitable for displaying in the explorer/experimenter gui

setSearchPercent

public void setSearchPercent(double p)

set the percentage of the search space to consider

Parameters:

p - percent of the search space ( 0 < p <= 100)

getSearchPercent

public double getSearchPercent()

get the percentage of the search space to consider

Returns:

the percent of the search space explored

getOptions

public java.lang.String[] getOptions()

Gets the current settings of RandomSearch.

Specified by:

getOptions in interface OptionHandler

Returns:

an array of strings suitable for passing to setOptions()

startSetToString

private java.lang.String startSetToString()

converts the array of starting attributes to a string. This is used by getOptions to return the actual attributes specified as the starting set. This is better than using m_startRanges.getRanges() as the same start set can be specified in different ways from the command line---eg 1,2,3 == 1-3. This is to ensure that stuff that is stored in a database is comparable.

Returns:

a comma seperated list of individual attribute numbers as a String

toString

public java.lang.String toString()

prints a description of the search

Returns:

a description of the search as a string

search

public int[] search(ASEvaluation ASEval,
                    Instances data)
             throws java.lang.Exception

Searches the attribute subset space randomly.

Specified by:

search in class ASSearch

Parameters:

data - the training instances.

ASEval - the attribute evaluator to guide the search

Returns:

an array (not necessarily ordered) of selected attribute indexes

Throws:

java.lang.Exception - if the search can't be completed

printSubset

private java.lang.String printSubset(java.util.BitSet temp)

prints a subset as a series of attribute numbers

Parameters:

temp - the subset to print

Returns:

a subset as a String of attribute numbers

attributeList

private int[] attributeList(java.util.BitSet group)

converts a BitSet into a list of attribute indexes

Parameters:

group - the BitSet to convert

Returns:

an array of attribute indexes

generateRandomSubset

private java.util.BitSet generateRandomSubset()

generates a random subset

Returns:

a random subset as a BitSet

countFeatures

private int countFeatures(java.util.BitSet featureSet)

counts the number of features in a subset

Parameters:

featureSet - the feature set for which to count the features

Returns:

the number of features in the subset

resetOptions

private void resetOptions()

resets to defaults