weka.distributed.hadoop

Class WekaClassifierEvaluationHadoopJob

java.lang.Object

distributed.core.DistributedJob

weka.distributed.hadoop.HadoopJob

weka.distributed.hadoop.WekaClassifierEvaluationHadoopJob

All Implemented Interfaces:

java.io.Serializable, CommandlineRunnable, EnvironmentHandler, OptionHandler, InstancesProducer, TextProducer

public class WekaClassifierEvaluationHadoopJob
extends HadoopJob
implements TextProducer, InstancesProducer, CommandlineRunnable

Hadoop job for running an evaluation of a classifier or regressor. Invokes up to four separate jobs (passes over the data). 1) Header creation, 2) optional randomly shuffled data chunk creation, 3) model construction and 4) model evaluation. Can perform evaluation on the training data, on a separate test set, or via cross-validation. In the case of the later, models for all folds are created and aggregated in one pass (job) and then evaluated on all folds in a second job.

Version:

$Revision: 10429 $

Author:

Mark Hall (mhall{[at]}pentaho{[dot]}com)

See Also:

Serialized Form

Nested Class Summary

Nested classes/interfaces inherited from class distributed.core.DistributedJob

distributed.core.DistributedJob.JobStatus

Field Summary

Fields inherited from class weka.distributed.hadoop.HadoopJob

COLT_JAR, DISTRIBUTED_WEKA_BASE_JAR, DISTRIBUTED_WEKA_HADOOP_JAR, JCOMMON_JAR, JFREECHART_JAR, LA4J_JAR, OPEN_CSV_JAR

Fields inherited from class distributed.core.DistributedJob

WEKA_ADDITIONAL_PACKAGES_KEY

Constructor Summary

Constructors

Constructor and Description
WekaClassifierEvaluationHadoopJob() Constructor

Method Summary

Methods

Modifier and Type	Method and Description
Instances	getInstances()
java.lang.String[]	getJobOptionsOnly() Get the options pertaining to this job only
int	getNumNodesInCluster() Get the number of nodes in the user's Hadoop cluster.
java.lang.String[]	getOptions()
java.lang.String	getSampleFractionForAUC() Get the percentage of predictions to retain (via uniform random sampling) for computing AUC and AUPRC.
java.lang.String	getSeparateTestSetPath() Get the path in HDFS to the separate test set to use.
java.lang.String	getText()
java.lang.String	globalInfo() Help info for this job
java.util.Enumeration<Option>	listOptions()
static void	main(java.lang.String[] args)
java.lang.String	numNodesInClusterTipText() Tip text for this property
void	run(java.lang.Object toRun, java.lang.String[] args)
boolean	runJob()
java.lang.String	sampleFractionForAUCTipText() Tip text for this property
java.lang.String	separateTestSetPathTipText() Tip text for this property
void	setNumNodesInCluster(int n) Set the number of nodes in the user's Hadoop cluster.
void	setOptions(java.lang.String[] options)
void	setSampleFractionForAUC(java.lang.String f) Set the percentage of predictions to retain (via uniform random sampling) for computing AUC and AUPRC.
void	setSeparateTestSetPath(java.lang.String path) Set the path in HDFS to the separate test set to use.

Methods inherited from class weka.distributed.hadoop.HadoopJob

additionalWekaPackagesTipText, cleanOutputDirectory, deubgTipText, getAdditionalWekaPackages, getBaseOptionsOnly, getDebug, getLoggingInterval, getMapNumber, getMapReduceJobConfig, getMapReduceNumber, getPathToWekaJar, getReduceNumber, loggingIntervalTipText, pathToWekaJarTipText, setAdditionalWekaPackages, setDebug, setLoggingInterval, setMapReduceJobConfig, setPathToWekaJar

Methods inherited from class distributed.core.DistributedJob

environmentSubstitute, getAdditionalWekaPackageNames, getJobName, getJobStatus, getLog, makeOptionsStr, parseInstance, setEnvironment, setJobDescription, setJobName, setJobStatus, setLog, setStatusMessagePrefix, stopJob

Methods inherited from class java.lang.Object

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

Constructor Detail

WekaClassifierEvaluationHadoopJob

public WekaClassifierEvaluationHadoopJob()

Constructor

Method Detail

globalInfo

public java.lang.String globalInfo()

Help info for this job

Returns:

help info for this job

listOptions

public java.util.Enumeration<Option> listOptions()

Specified by:

listOptions in interface OptionHandler

Overrides:

listOptions in class HadoopJob

setOptions

public void setOptions(java.lang.String[] options)
                throws java.lang.Exception

Specified by:

setOptions in interface OptionHandler

Overrides:

setOptions in class HadoopJob

Throws:

java.lang.Exception

getOptions

public java.lang.String[] getOptions()

Specified by:

getOptions in interface OptionHandler

Overrides:

getOptions in class HadoopJob

getJobOptionsOnly

public java.lang.String[] getJobOptionsOnly()

Get the options pertaining to this job only

Returns:

the options for this job only

separateTestSetPathTipText

public java.lang.String separateTestSetPathTipText()

Tip text for this property

Returns:

the tip text for this property

setSeparateTestSetPath

public void setSeparateTestSetPath(java.lang.String path)

Set the path in HDFS to the separate test set to use. Either this or the total number of folds should be specified (but not both).

Parameters:

path - the path in HDFS to the separate test set to evaluate on

getSeparateTestSetPath

public java.lang.String getSeparateTestSetPath()

Get the path in HDFS to the separate test set to use. Either this or the total number of folds should be specified (but not both).

Returns:

the path in HDFS to the separate test set to evaluate on

sampleFractionForAUCTipText

public java.lang.String sampleFractionForAUCTipText()

Tip text for this property

Returns:

the tip text for this property

setSampleFractionForAUC

public void setSampleFractionForAUC(java.lang.String f)

Set the percentage of predictions to retain (via uniform random sampling) for computing AUC and AUPRC. If not specified, then no predictions are retained and these metrics are not computed.

Parameters:

f - the fraction (between 0 and 1) of all predictions to retain for computing AUC/AUPRC.

getSampleFractionForAUC

public java.lang.String getSampleFractionForAUC()

Get the percentage of predictions to retain (via uniform random sampling) for computing AUC and AUPRC. If not specified, then no predictions are retained and these metrics are not computed.

Returns:

the fraction (between 0 and 1) of all predictions to retain for computing AUC/AUPRC.

numNodesInClusterTipText

public java.lang.String numNodesInClusterTipText()

Tip text for this property

Returns:

the tip text for this property

setNumNodesInCluster

public void setNumNodesInCluster(int n)

Set the number of nodes in the user's Hadoop cluster. Used for setting the number of reducers to use when performing a cross-validation.

Parameters:

n - the number of nodes in the cluster

getNumNodesInCluster

public int getNumNodesInCluster()

Get the number of nodes in the user's Hadoop cluster. Used for setting the number of reducers to use when performing a cross-validation.

Returns:

the number of nodes in the cluster

runJob

public boolean runJob()
               throws weka.distributed.DistributedWekaException

Specified by:

runJob in class distributed.core.DistributedJob

Throws:

weka.distributed.DistributedWekaException

getText

public java.lang.String getText()

Specified by:

getText in interface TextProducer

getInstances

public Instances getInstances()

Specified by:

getInstances in interface InstancesProducer

main

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

run

public void run(java.lang.Object toRun,
       java.lang.String[] args)
         throws java.lang.IllegalArgumentException

Specified by:

run in interface CommandlineRunnable

Throws:

java.lang.IllegalArgumentException