All Packages  Class Hierarchy  This Package  Previous  Next  Index  WEKA's home

Class weka.classifiers.rules.JRip 

java.lang.Object
    |
    +----weka.classifiers.Classifier
            |
            +----weka.classifiers.DistributionClassifier
                    |
                    +----weka.classifiers.rules.JRip
public class JRip
extends DistributionClassifier
implements OptionHandler, AdditionalMeasureProducer, WeightedInstancesHandler
This class implements a propositional rule learner, Repeated Incremental Pruning to Produce Error Reduction (RIPPER), which is proposed by William W. Cohen as an optimzed version of IREP.

The algorithm is briefly described as follows:

Initialize RS = {}, and for each class from the less prevalent one to the more frequent one, DO:

1. Building stage: repeat 1.1 and 1.2 until the descrition length (DL) of the ruleset and examples is 64 bits greater than the smallest DL met so far, or there are no positive examples, or the error rate >= 50%.

1.1. Grow phase:
Grow one rule by greedily adding antecedents (or conditions) to the rule until the rule is perfect (i.e. 100% accurate). The procedure tries every possible value of each attribute and selects the condition with highest information gain: p(log(p/t)-log(P/T)).

1.2. Prune phase:
Incrementally prune each rule and allow the pruning of any final sequences of the antecedents;
The pruning metric is (p-n)/(p+n) -- but it's actually 2p/(p+n) -1, so in this implementation we simply use p/(p+n) (actually (p+1)/(p+n+2), thus if p+n is 0, it's 0.5).

2. Optimization stage: after generating the initial ruleset {Ri}, generate and prune two variants of each rule Ri from randomized data using procedure 1.1 and 1.2. But one variant is generated from an empty rule while the other is generated by greedily adding antecedents to the original rule. Moreover, the pruning metric used here is (TP+TN)/(P+N).
Then the smallest possible DL for each variant and the original rule is computed. The variant with the minimal DL is selected as the final representative of Ri in the ruleset.
After all the rules in {Ri} have been examined and if there are still residual positives, more rules are generated based on the residual positives using Building Stage again.

3. Delete the rules from the ruleset that would increase the DL of the whole ruleset if it were in it. and add resultant ruleset to RS.

ENDDO

Note that there seem to be 2 bugs in the ripper program that would affect the ruleset size and accuracy slightly. This implementation avoids these bugs and thus is a little bit different from Cohen's original implementation. Even after fixing the bugs, since the order of classes with the same frequency is not defined in ripper, there still seems to be some trivial difference between this implementation and the original ripper, especially for audiology data in UCI repository, where there are lots of classes of few instances.

If wrapped by other classes, typical usage of this class is:
JRip rip = new JRip(); Instances data = ... // Data from somewhere double[] orderedClasses = ... // Get the ordered class counts for the data double expFPRate = ... // Calculate the expected FP/(FP+FN) rate double classIndex = ... // The class index for which ruleset is built // DL of default rule, no theory DL, only data DL double defDL = RuleStats.dataDL(expFPRate, 0.0, data.sumOfWeights(), 0.0, orderedClasses[(int)classIndex]); rip.rulesetForOneClass(expFPRate, data, classIndex, defDL); RuleStats rulesetStats = rip.getRuleStats(0); // Can get heaps of information from RuleStats, e.g. combined DL, // simpleStats, etc. double comDL = rulesetStats.combinedDL(expFPRate, classIndex); int whichRule = ... // Want simple stats of which rule? double[] simpleStats = rulesetStats.getSimpleStats(whichRule); ... Details please see "Fast Effective Rule Induction", William W. Cohen, 'Machine Learning: Proceedings of the Twelfth International Conference' (ML95).

PS. We have compared this implementation with the original ripper implementation in aspects of accuracy, ruleset size and running time on both artificial data "ab+bcd+defg" and UCI datasets. In all these aspects it seems to be quite comparable to the original ripper implementation. However, we didn't consider memory consumption optimization in this implementation.

Version:
$Revision: 1.9 $
Author:
Xin Xu (xx5@cs.waikato.ac.nz)
Author:
Eibe Frank (eibe@cs.waikato.ac.nz)

Constructor Index

o JRip()
 

Method Index

o buildClassifier(Instances)
Builds Ripper in the order of class frequencies.
o distributionForInstance(Instance)
Classify the test instance with the rule learner and provide the class distributions
o enumerateMeasures()
Returns an enumeration of the additional measure names
o getCheckErrorRate()
 
o getDebug()
 
o getFolds()
 
o getMeasure(String)
Returns the value of the named measure
o getMinNo()
 
o getOptimizations()
 
o getOptions()
Gets the current settings of the Classifier.
o getRuleset()
Get the ruleset generated by Ripper
o getRuleStats(int)
Get the statistics of the ruleset in the given position
o getSeed()
 
o getUsePruning()
 
o listOptions()
Returns an enumeration describing the available options Valid options are:

-F number
The number of folds for reduced error pruning.

o main(String[])
Main method.
o setCheckErrorRate(boolean)
 
o setDebug(boolean)
 
o setFolds(int)
 
o setMinNo(double)
 
o setOptimizations(int)
 
o setOptions(String[])
Parses a given list of options.
o setSeed(long)
 
o setUsePruning(boolean)
 
o toString()
Prints the all the rules of the rule learner.

Constructor Detail

o JRip
public JRip()

Method Detail

o listOptions
public java.util.Enumeration listOptions()
          Returns an enumeration describing the available options Valid options are:

-F number
The number of folds for reduced error pruning. One fold is used as the pruning set. (Default: 3)

-N number
The minimal weights of instances within a split. (Default: 2)

-O number
Set the number of runs of optimizations. (Default: 2)

-D
Whether turn on the debug mode -S number
The seed of randomization used in Ripper.(Default: 1)

-E
Whether NOT check the error rate >= 0.5 in stopping criteria. (default: check)

-P
Whether NOT use pruning. (default: use pruning)

Returns:
an enumeration of all the available options
o setOptions
public void setOptions(java.lang.String options[]) throws java.lang.Exception
          Parses a given list of options.
Parameters:
options - the list of options as an array of strings
Throws:
java.lang.Exception - if an option is not supported
o getOptions
public java.lang.String[] getOptions()
          Gets the current settings of the Classifier.
Returns:
an array of strings suitable for passing to setOptions
o enumerateMeasures
public java.util.Enumeration enumerateMeasures()
          Returns an enumeration of the additional measure names
Returns:
an enumeration of the measure names
o getMeasure
public double getMeasure(java.lang.String additionalMeasureName)
          Returns the value of the named measure
Parameters:
measureName - the name of the measure to query for its value
Returns:
the value of the named measure
Throws:
java.lang.IllegalArgumentException - if the named measure is not supported
o setFolds
public void setFolds(int fold)
o getFolds
public int getFolds()
o setMinNo
public void setMinNo(double m)
o getMinNo
public double getMinNo()
o setSeed
public void setSeed(long s)
o getSeed
public long getSeed()
o setOptimizations
public void setOptimizations(int run)
o getOptimizations
public int getOptimizations()
o setDebug
public void setDebug(boolean d)
o getDebug
public boolean getDebug()
o setCheckErrorRate
public void setCheckErrorRate(boolean d)
o getCheckErrorRate
public boolean getCheckErrorRate()
o setUsePruning
public void setUsePruning(boolean d)
o getUsePruning
public boolean getUsePruning()
o getRuleset
public FastVector getRuleset()
          Get the ruleset generated by Ripper
Returns:
the ruleset
o getRuleStats
public RuleStats getRuleStats(int pos)
          Get the statistics of the ruleset in the given position
Parameters:
pos - the position of the stats, assuming correct
o buildClassifier
public void buildClassifier(Instances instances) throws java.lang.Exception
          Builds Ripper in the order of class frequencies. For each class it's built in two stages: building and optimization
Parameters:
instances - the training data
Throws:
java.lang.Exception - if classifier can't be built successfully
Overrides:
buildClassifier in class Classifier
o distributionForInstance
public double[] distributionForInstance(Instance datum)
          Classify the test instance with the rule learner and provide the class distributions
Parameters:
datum - the instance to be classified
Returns:
the distribution
Overrides:
distributionForInstance in class DistributionClassifier
o toString
public java.lang.String toString()
          Prints the all the rules of the rule learner.
Returns:
a textual description of the classifier
Overrides:
toString in class java.lang.Object
o main
public static void main(java.lang.String args[])
          Main method.
Parameters:
args - the options for the classifier 
All Packages  Class Hierarchy  This Package  Previous  Next  Index  WEKA's home