Package weka.classifiers.evaluation

Class AbstractEvaluationMetric

java.lang.Object

weka.classifiers.evaluation.AbstractEvaluationMetric

All Implemented Interfaces:

Serializable

public abstract class AbstractEvaluationMetric
extends Object
implements Serializable

Abstract base class for pluggable classification/regression evaluation metrics.

Version:

$Revision$

Author:

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

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
class	AbstractEvaluationMetric.UnknownStatisticException	Exception for subclasses to throw if asked for a statistic that is not part of their implementation

Constructor Summary

Constructors

Constructor	Description
AbstractEvaluationMetric()

Method Summary

Modifier and Type	Method	Description
abstract boolean	appliesToNominalClass()	Return true if this evaluation metric can be computed when the class is nominal
abstract boolean	appliesToNumericClass()	Return true if this evaluation metric can be computed when the class is numeric
abstract String	getMetricDescription()	Get a short description of this metric (algorithm, forumulas etc.).
abstract String	getMetricName()	Get the name of this metric
static ArrayList<AbstractEvaluationMetric>	getPluginMetrics()	Gets a list of freshly instantiated concrete implementations of available plugin metrics or null if there are no plugin metrics available
abstract double	getStatistic(String statName)	Get the value of the named statistic
abstract List<String>	getStatisticNames()	Get a list of the names of the statistics that this metrics computes.
void	setBaseEvaluation(Evaluation eval)	Set the base evaluation object to use.
boolean	statisticIsMaximisable(String statName)	True if the optimum value of the named metric is a maximum value; false if the optimim value is a minimum value.

Methods inherited from class java.lang.Object

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

Constructor Detail

AbstractEvaluationMetric

public AbstractEvaluationMetric()

Method Detail

getPluginMetrics

public static ArrayList<AbstractEvaluationMetric> getPluginMetrics()

Gets a list of freshly instantiated concrete implementations of available plugin metrics or null if there are no plugin metrics available

Returns:

a list of plugin metrics or null if there are no plugin metrics

setBaseEvaluation

public void setBaseEvaluation(Evaluation eval)

Set the base evaluation object to use. IMPORTANT: subclasses should treat this object as read-only.

Parameters:

eval -

appliesToNominalClass

public abstract boolean appliesToNominalClass()

Return true if this evaluation metric can be computed when the class is nominal

Returns:

true if this evaluation metric can be computed when the class is nominal

appliesToNumericClass

public abstract boolean appliesToNumericClass()

Return true if this evaluation metric can be computed when the class is numeric

Returns:

true if this evaluation metric can be computed when the class is numeric

getMetricName

public abstract String getMetricName()

Get the name of this metric

Returns:

the name of this metric

getMetricDescription

public abstract String getMetricDescription()

Get a short description of this metric (algorithm, forumulas etc.).

Returns:

a short description of this metric

getStatisticNames

public abstract List<String> getStatisticNames()

Get a list of the names of the statistics that this metrics computes. E.g. an information theoretic evaluation measure might compute total number of bits as well as average bits/instance

Returns:

the names of the statistics that this metric computes

getStatistic

public abstract double getStatistic(String statName)

Get the value of the named statistic

Parameters:

statName - the name of the statistic to compute the value for

Returns:

the computed statistic or Utils.missingValue() if the statistic can't be computed for some reason

statisticIsMaximisable

public boolean statisticIsMaximisable(String statName)

True if the optimum value of the named metric is a maximum value; false if the optimim value is a minimum value. Subclasses should override this method to suit their statistic(s)

Returns:

true (default implementation)