weka.classifiers.rules
Class ConjunctiveRule.NumericAntd

java.lang.Object
  weka.classifiers.rules.ConjunctiveRule.Antd
      weka.classifiers.rules.ConjunctiveRule.NumericAntd

All Implemented Interfaces:

java.io.Serializable

Enclosing class:

ConjunctiveRule

private class ConjunctiveRule.NumericAntd

extends ConjunctiveRule.Antd

The antecedent with numeric attribute

Field Summary

protected Attribute	att The attribute of the antecedent
protected double	inform The information of this antecedent test on the growing data
protected double	maxInfoGain The maximum infoGain achieved by this antecedent test
private double	splitPoint
protected double[]	uncover The parameters related to the data not covered by the previous antecedents when the class is nominal
protected double	uncoverSum The parameter related to the meanSquaredError of the data not covered by the previous antecedents when the class is numeric
protected double	uncoverWtSq The parameter related to the meanSquaredError of the data not covered by the previous antecedents when the class is numeric
protected double	uncoverWtVl The parameter related to the meanSquaredError of the data not covered by the previous antecedents when the class is numeric
protected double	value The attribute value of the antecedent.

Constructor Summary

ConjunctiveRule.NumericAntd(Attribute a, double[] unc)

ConjunctiveRule.NumericAntd(Attribute a, double sq, double vl, double wts)

Method Summary

protected double	entropy(double[] value, double sum) Function used to calculate the entropy of given vector of values entropy = (1/sum)*{-sigma[i=1..P](Xi*log2(Xi)) + sum*log2(sum)} where P is the length of the vector
Attribute	getAttr()
double	getAttrValue()
double	getInfo()
double	getMaxInfoGain()
double	getSplitPoint()
boolean	isCover(Instance inst) Whether the instance is covered by this antecedent
Instances[]	splitData(Instances insts, double defInfo) Implements the splitData function.
java.lang.String	toString() Prints this antecedent
protected double	wtMeanSqErr(double weightedSq, double weightedValue, double sum) Function used to calculate the weighted mean squared error, i.e., sum[x-avg(x)]^2 based on the given elements of the formula: meanSquaredError = sum(Wi*Xi^2) - (sum(WiXi))^2/sum(Wi)

Methods inherited from class java.lang.Object

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

Field Detail

splitPoint

private double splitPoint

att

protected Attribute att

The attribute of the antecedent

value

protected double value

The attribute value of the antecedent. For numeric attribute, value is either 0(1st bag) or 1(2nd bag)

maxInfoGain

protected double maxInfoGain

The maximum infoGain achieved by this antecedent test

inform

protected double inform

The information of this antecedent test on the growing data

uncoverWtSq

protected double uncoverWtSq

The parameter related to the meanSquaredError of the data not covered by the previous antecedents when the class is numeric

uncoverWtVl

protected double uncoverWtVl

The parameter related to the meanSquaredError of the data not covered by the previous antecedents when the class is numeric

uncoverSum

protected double uncoverSum

The parameter related to the meanSquaredError of the data not covered by the previous antecedents when the class is numeric

uncover

protected double[] uncover

The parameters related to the data not covered by the previous antecedents when the class is nominal

Constructor Detail

ConjunctiveRule.NumericAntd

public ConjunctiveRule.NumericAntd(Attribute a,
                                   double[] unc)

ConjunctiveRule.NumericAntd

public ConjunctiveRule.NumericAntd(Attribute a,
                                   double sq,
                                   double vl,
                                   double wts)

Method Detail

getSplitPoint

public double getSplitPoint()

splitData

public Instances[] splitData(Instances insts,
                             double defInfo)

Implements the splitData function. This procedure is to split the data into two bags according to the information gain of the numeric attribute value the data with missing values are stored in the last split. The maximum infoGain is also calculated.

Specified by:

splitData in class ConjunctiveRule.Antd

Parameters:

insts - the data to be split

defInfo - the default information for data

Returns:

the array of data after split

isCover

public boolean isCover(Instance inst)

Whether the instance is covered by this antecedent

Specified by:

isCover in class ConjunctiveRule.Antd

Parameters:

inst - the instance in question

Returns:

the boolean value indicating whether the instance is covered by this antecedent

toString

public java.lang.String toString()

Prints this antecedent

Specified by:

toString in class ConjunctiveRule.Antd

Returns:

a textual description of this antecedent

getAttr

public Attribute getAttr()

getAttrValue

public double getAttrValue()

getMaxInfoGain

public double getMaxInfoGain()

getInfo

public double getInfo()

wtMeanSqErr

protected double wtMeanSqErr(double weightedSq,
                             double weightedValue,
                             double sum)

Function used to calculate the weighted mean squared error, i.e., sum[x-avg(x)]^2 based on the given elements of the formula: meanSquaredError = sum(Wi*Xi^2) - (sum(WiXi))^2/sum(Wi)

Parameters:

weightedSq - sum(Wi*Xi^2)

weightedValue - sum(WiXi)

sum - sum of weights

Returns:

the weighted mean-squared error

entropy

protected double entropy(double[] value,
                         double sum)

Function used to calculate the entropy of given vector of values entropy = (1/sum)*{-sigma[i=1..P](Xi*log2(Xi)) + sum*log2(sum)} where P is the length of the vector

Parameters:

value - the given vector of values

sum - the sum of the given values. It's provided just for efficiency.

Returns:

the entropy