weka.classifiers
Class VFI

java.lang.Object
  |
  +--weka.classifiers.Classifier
        |
        +--weka.classifiers.DistributionClassifier
              |
              +--weka.classifiers.VFI

All Implemented Interfaces:

java.lang.Cloneable, OptionHandler, java.io.Serializable, WeightedInstancesHandler

public class VFI

extends DistributionClassifier

implements OptionHandler, WeightedInstancesHandler

Class implementing the voting feature interval classifier. For numeric attributes, upper and lower boundaries (intervals) are constructed around each class. Discrete attributes have point intervals. Class counts are recorded for each interval on each feature. Classification is by voting. Missing values are ignored. Does not handle numeric class.

Have added a simple attribute weighting scheme. Higher weight is assigned to more confident intervals, where confidence is a function of entropy: weight (att_i) = (entropy of class distrib att_i / max uncertainty)^-bias.

Faster than NaiveBayes but slower than HyperPipes.

  Confidence: 0.01 (two tailed)

 Dataset                   (1) VFI '-B  | (2) Hyper (3) Naive
                         ------------------------------------
 anneal.ORIG               (10)   74.56 |   97.88 v   74.77
 anneal                    (10)   71.83 |   97.88 v   86.51 v
 audiology                 (10)   51.69 |   66.26 v   72.25 v
 autos                     (10)   57.63 |   62.79 v   57.76
 balance-scale             (10)   68.72 |   46.08 *   90.5  v
 breast-cancer             (10)   67.25 |   69.84 v   73.12 v
 wisconsin-breast-cancer   (10)   95.72 |   88.31 *   96.05 v
 horse-colic.ORIG          (10)   66.13 |   70.41 v   66.12
 horse-colic               (10)   78.36 |   62.07 *   78.28
 credit-rating             (10)   85.17 |   44.58 *   77.84 *
 german_credit             (10)   70.81 |   69.89 *   74.98 v
 pima_diabetes             (10)   62.13 |   65.47 v   75.73 v
 Glass                     (10)   56.82 |   50.19 *   47.43 *
 cleveland-14-heart-diseas (10)   80.01 |   55.18 *   83.83 v
 hungarian-14-heart-diseas (10)   82.8  |   65.55 *   84.37 v
 heart-statlog             (10)   79.37 |   55.56 *   84.37 v
 hepatitis                 (10)   83.78 |   63.73 *   83.87
 hypothyroid               (10)   92.64 |   93.33 v   95.29 v
 ionosphere                (10)   94.16 |   35.9  *   82.6  *
 iris                      (10)   96.2  |   91.47 *   95.27 *
 kr-vs-kp                  (10)   88.22 |   54.1  *   87.84 *
 labor                     (10)   86.73 |   87.67     93.93 v
 lymphography              (10)   78.48 |   58.18 *   83.24 v
 mushroom                  (10)   99.85 |   99.77 *   95.77 *
 primary-tumor             (10)   29    |   24.78 *   49.35 v
 segment                   (10)   77.42 |   75.15 *   80.1  v
 sick                      (10)   65.92 |   93.85 v   92.71 v
 sonar                     (10)   58.02 |   57.17     67.97 v
 soybean                   (10)   86.81 |   86.12 *   92.9  v
 splice                    (10)   88.61 |   41.97 *   95.41 v
 vehicle                   (10)   52.94 |   32.77 *   44.8  *
 vote                      (10)   91.5  |   61.38 *   90.19 *
 vowel                     (10)   57.56 |   36.34 *   62.81 v
 waveform                  (10)   56.33 |   46.11 *   80.02 v
 zoo                       (10)   94.05 |   94.26     95.04 v
                          ------------------------------------
                                (v| |*) |  (9|3|23)  (22|5|8) 
 

For more information, see

Demiroz, G. and Guvenir, A. (1997) "Classification by voting feature intervals", ECML-97.

Valid options are:

-C
Don't Weight voting intervals by confidence.

-B
Set exponential bias towards confident intervals. default = 1.0

Author:

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

See Also:

Serialized Form

Field Summary

protected double	m_bias Bias towards more confident intervals
protected int	m_ClassIndex The index of the class attribute
protected double[][][]	m_counts The class counts for each interval of each attribute
protected double[]	m_globalCounts The global class counts
protected Instances	m_Instances The training data
protected double[][]	m_intervalBounds The lower bounds for each attribute
protected double	m_maxEntrop The maximum entropy for the class
protected int	m_NumClasses The number of classes
protected boolean	m_weightByConfidence Exponentially bias more confident intervals

Constructor Summary

VFI()

Method Summary

java.lang.String	biasTipText() Returns the tip text for this property
void	buildClassifier(Instances instances) Generates the classifier.
double[]	distributionForInstance(Instance instance) Classifies the given test instance.
double	getBias() Get the value of the bias parameter
java.lang.String[]	getOptions() Gets the current settings of VFI
boolean	getWeightByConfidence() Get whether feature intervals are being weighted by confidence
java.lang.String	globalInfo() Returns a string describing this search method
java.util.Enumeration	listOptions() Returns an enumeration describing the available options
static void	main(java.lang.String[] args) Main method for testing this class.
void	setBias(double b) Set the value of the exponential bias towards more confident intervals
void	setOptions(java.lang.String[] options) Parses a given list of options.
void	setWeightByConfidence(boolean c) Set weighting by confidence
java.lang.String	toString() Returns a description of this classifier.
java.lang.String	weightByConfidenceTipText() Returns the tip text for this property

Methods inherited from class weka.classifiers.DistributionClassifier

classifyInstance

Methods inherited from class weka.classifiers.Classifier

forName, makeCopies

Methods inherited from class java.lang.Object

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

Field Detail

m_ClassIndex

protected int m_ClassIndex

The index of the class attribute

m_NumClasses

protected int m_NumClasses

The number of classes

m_Instances

protected Instances m_Instances

The training data

m_counts

protected double[][][] m_counts

The class counts for each interval of each attribute

m_globalCounts

protected double[] m_globalCounts

The global class counts

m_intervalBounds

protected double[][] m_intervalBounds

The lower bounds for each attribute

m_maxEntrop

protected double m_maxEntrop

The maximum entropy for the class

m_weightByConfidence

protected boolean m_weightByConfidence

Exponentially bias more confident intervals

m_bias

protected double m_bias

Bias towards more confident intervals

Constructor Detail

VFI

public VFI()

Method Detail

globalInfo

public java.lang.String globalInfo()

Returns a string describing this search method

Returns:

a description of the search method 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:

-C
Don't weight voting intervals by confidence.

-B
Set exponential bias towards confident intervals. default = 1.0

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

weightByConfidenceTipText

public java.lang.String weightByConfidenceTipText()

Returns the tip text for this property

Returns:

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

setWeightByConfidence

public void setWeightByConfidence(boolean c)

Set weighting by confidence

Parameters:

c - true if feature intervals are to be weighted by confidence

getWeightByConfidence

public boolean getWeightByConfidence()

Get whether feature intervals are being weighted by confidence

Returns:

true if weighting by confidence is selected

biasTipText

public java.lang.String biasTipText()

Returns the tip text for this property

Returns:

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

setBias

public void setBias(double b)

Set the value of the exponential bias towards more confident intervals

Parameters:

b - the value of the bias parameter

getBias

public double getBias()

Get the value of the bias parameter

Returns:

the bias parameter

getOptions

public java.lang.String[] getOptions()

Gets the current settings of VFI

Specified by:

getOptions in interface OptionHandler

Returns:

an array of strings suitable for passing to setOptions()

buildClassifier

public void buildClassifier(Instances instances)
                     throws java.lang.Exception

Generates the classifier.

Overrides:

buildClassifier in class Classifier

Parameters:

instances - set of instances serving as training data

Throws:

java.lang.Exception - if the classifier has not been generated successfully

toString

public java.lang.String toString()

Returns a description of this classifier.

Overrides:

toString in class java.lang.Object

Returns:

a description of this classifier as a string.

distributionForInstance

public double[] distributionForInstance(Instance instance)
                                 throws java.lang.Exception

Classifies the given test instance.

Overrides:

distributionForInstance in class DistributionClassifier

Parameters:

instance - the instance to be classified

Returns:

the predicted class for the instance

Throws:

java.lang.Exception - if the instance can't be classified

main

public static void main(java.lang.String[] args)

Main method for testing this class.

Parameters:

args - should contain command line arguments for evaluation (see Evaluation).