uk.ac.ed.inf.enzml.mulan.attributesfilter
Class AttributesFilter

java.lang.Object
  uk.ac.ed.inf.utils.Initialised
      uk.ac.ed.inf.enzml.mulan.attributesfilter.AttributesFilter

All Implemented Interfaces:

uk.ac.ed.inf.utils.Initialisable

public class AttributesFilter
extends uk.ac.ed.inf.utils.Initialised

This class filters the attributes of a dataset (the test set) to make it compatible with another dataset (the train set). The result is a new dataset that can be used as test set for the specified train set. Main operations: 1. Attributes: only attributes present in the train set can appear in the test set (& all of them must be present) 2. Class attributes: all the classes of the train set *and* the test set must be conserved, to allow for train/test evaluation. This means that also the train arff must be regenerated to include all the classes present in the test set (the class numbering will change). 3. Mulan XML file (class labels): cannot be the same XML file of the training set, must also include all the classes in the test set. 4. New test ARFF File: the new test set arff file name should mention the original files and which attributes and classes were used for filtering (that is, left in the file or added to the file). 5. New train ARFF file: (including also the test classes) 'trainset id' + 'including_classes_of' + 'test_set id' + '.arff' 6. New test ARFF record: when stored in the database, should contain trace (id/props file path) of the original training set and the "old" test set used to generate the "new", filtered, test set.

Version:

3 Sep 2010

Author:

Luna De Ferrari luna.deferrari-at-ed.ac.uk

Field Summary

Fields inherited from class uk.ac.ed.inf.utils.Initialised

m_initialisationIsCorrect

Constructor Summary

AttributesFilter(int oldTrainArffId, int oldTestArffId, java.lang.String dataDatabaseProps, java.lang.String arffAndResultsPropsFileName)

Method Summary

int	createNewArff(int oldArffRecordId)
void	createNewTestArff()
void	createNewTrainArff()
java.lang.String	getArffAndResultsProperties()
uk.ac.ed.inf.utils.database.DbManager	getArffRecordsDbManager()
ArffPropsTableReader	getArffTableReader()
uk.ac.ed.inf.utils.database.DbManager	getDataDbManager()
java.lang.String	getDataDbPath()
int	getNewTestArffId()
int	getNewTrainArffId()
int	getOldTestArffId()
int	getOldTrainArffId() public MulanArff getTrainArff() { return m_trainArff; }
java.util.TreeSet<java.lang.String>	getTrainAttributesAndTrainTestClasses()
java.lang.String	getTrainXmlPath()
java.lang.String	getXmlPathById(int arffId)
java.lang.String	getXmlQueryById(int arffId)
static void	main(java.lang.String[] args)
static int[]	run(int oldTrainArffId, int oldTestArffId, java.lang.String dataDbPropsFile, java.lang.String resultsPropsFileName) Filter the old test arff by the train arff given and regenerate the old train set using all train+test classes.

Methods inherited from class uk.ac.ed.inf.utils.Initialised

initialisationIsCorrect, setInitialisationIsCorrect

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

AttributesFilter

public AttributesFilter(int oldTrainArffId,
                        int oldTestArffId,
                        java.lang.String dataDatabaseProps,
                        java.lang.String arffAndResultsPropsFileName)

Method Detail

createNewArff

public int createNewArff(int oldArffRecordId)

Parameters:

oldArffRecordId - the database id of the old arff file

Returns:

the database id of the new, filtered arff file

createNewTestArff

public void createNewTestArff()

createNewTrainArff

public void createNewTrainArff()

getArffAndResultsProperties

public java.lang.String getArffAndResultsProperties()

getArffRecordsDbManager

public uk.ac.ed.inf.utils.database.DbManager getArffRecordsDbManager()

getArffTableReader

public ArffPropsTableReader getArffTableReader()

getDataDbManager

public uk.ac.ed.inf.utils.database.DbManager getDataDbManager()

getDataDbPath

public java.lang.String getDataDbPath()

getNewTestArffId

public int getNewTestArffId()

getNewTrainArffId

public int getNewTrainArffId()

getOldTestArffId

public int getOldTestArffId()

getOldTrainArffId

public int getOldTrainArffId()

public MulanArff getTrainArff() { return m_trainArff; }

getTrainAttributesAndTrainTestClasses

public java.util.TreeSet<java.lang.String> getTrainAttributesAndTrainTestClasses()

getTrainXmlPath

public java.lang.String getTrainXmlPath()

getXmlPathById

public java.lang.String getXmlPathById(int arffId)

getXmlQueryById

public java.lang.String getXmlQueryById(int arffId)

main

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

run

public static int[] run(int oldTrainArffId,
                        int oldTestArffId,
                        java.lang.String dataDbPropsFile,
                        java.lang.String resultsPropsFileName)

Filter the old test arff by the train arff given and regenerate the old train set using all train+test classes.

Parameters:

oldTrainArffId - the record id of the original training arff

oldTestArffId - the record id of the original test arff

resultsPropsFileName - the database where to write the new arff records

dataDbPropsFile - the database where to read the data to rebuild the arff files

Returns:

the identifiers of the new train and test arff records in the database