weka.classifiers.rules
Class Ridor.Ridor_node

java.lang.Object
  weka.classifiers.rules.Ridor.Ridor_node

All Implemented Interfaces:

java.io.Serializable

Enclosing class:

Ridor

private class Ridor.Ridor_node

extends java.lang.Object

implements java.io.Serializable

Private class implementing the single node of Ridor. It consists of a default class label, a set of exceptions to the default rule and the exceptions to each exception

Field Summary

private double	defClass The default class label
private Ridor.Ridor_node[]	excepts The exceptions of the exception rules
private int	level The level of this node
private Ridor.RidorRule[]	rules The set of exceptions of the default rule.

Constructor Summary

private

Ridor.Ridor_node()

Method Summary

private Instances	append(Instances data1, Instances data2) Private function to combine two data
private double	buildRuleset(Instances insts, double classCount, java.util.Vector ruleset) Private function to build a rule set and return the weighted avg of accuracy rate of rules in the set.
private double	computeWeightedAcRt(double worthRt, double cover, double total) Compute the weighted average of accuracy rate of a certain rule Each rule is weighted by its coverage proportion in the whole data.
private Instances[][]	divide(Ridor.RidorRule rule, Instances[] dataByClass) Builds an array of data according to their true class label Each bag of data is filtered through the rule specified and is totally covered by this rule.
void	findRules(Instances[] dataByClass, int lvl) Builds a ripple-down manner rule learner.
double	getDefClass() "Get" member functions
Ridor.Ridor_node[]	getExcepts()
Ridor.RidorRule[]	getRules()
int	size() The size of the certain node of Ridor, i.e. the number of rules generated within and below this node
java.lang.String	toString() Prints the all the rules of one node of Ridor.

Methods inherited from class java.lang.Object

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

Field Detail

defClass

private double defClass

The default class label

rules

private Ridor.RidorRule[] rules

The set of exceptions of the default rule. Each element also has its own exceptions and the consequent of each rule is determined by its exceptions

excepts

private Ridor.Ridor_node[] excepts

The exceptions of the exception rules

level

private int level

The level of this node

Constructor Detail

Ridor.Ridor_node

private Ridor.Ridor_node()

Method Detail

getDefClass

public double getDefClass()

"Get" member functions

getRules

public Ridor.RidorRule[] getRules()

getExcepts

public Ridor.Ridor_node[] getExcepts()

findRules

public void findRules(Instances[] dataByClass,
                      int lvl)
               throws java.lang.Exception

Builds a ripple-down manner rule learner.

Parameters:

dataByClass - the divided data by their class label. The real class labels of the instances are all set to 0

lvl - the level of the parent node

Throws:

java.lang.Exception - if ruleset of this node cannot be built

buildRuleset

private double buildRuleset(Instances insts,
                            double classCount,
                            java.util.Vector ruleset)
                     throws java.lang.Exception

Private function to build a rule set and return the weighted avg of accuracy rate of rules in the set.

Parameters:

insts - the data used to build ruleset

classCount - the counts of the instances with the predicted class but not yet covered by the ruleset

ruleset - the ruleset to be built

Returns:

the weighted accuracy rate of the ruleset

Throws:

if - the rules cannot be built properly

java.lang.Exception

append

private Instances append(Instances data1,
                         Instances data2)

Private function to combine two data

Parameters:

data1 - the data to which data2 is appended

data2 - the data to be appended to data1

Returns:

the merged data

computeWeightedAcRt

private double computeWeightedAcRt(double worthRt,
                                   double cover,
                                   double total)

Compute the weighted average of accuracy rate of a certain rule Each rule is weighted by its coverage proportion in the whole data. So the accuracy rate of one ruleset is actually (worth rate) * (coverage proportion) coverage of the rule on the whole data where coverage proportion = ----------------------------------------- the whole data size fed into the ruleset

Parameters:

worthRt - the worth rate

cover - the coverage of the rule on the whole data

total - the total data size fed into the ruleset

Returns:

the weighted accuracy rate of this rule

divide

private Instances[][] divide(Ridor.RidorRule rule,
                             Instances[] dataByClass)

Builds an array of data according to their true class label Each bag of data is filtered through the rule specified and is totally covered by this rule. Both the data covered and uncovered by the rule will be returned by the procedure.

Parameters:

rule - the rule covering the data

dataByClass - the array of data to be covered by the rule

Returns:

the arrays of data both covered and not covered by the rule

size

public int size()

The size of the certain node of Ridor, i.e. the number of rules generated within and below this node

Returns:

the size of this node

toString

public java.lang.String toString()

Prints the all the rules of one node of Ridor.

Returns:

a textual description of one node of Ridor