org.dmilne.weka.wrapper

Class Decider<A extends Enum<A>,C>

java.lang.Object

org.dmilne.weka.wrapper.Decider<A,C>

Type Parameters:

A - An enumeration of attributes (features) used to inform the decision function

C - The expected output (this must either be Double, Boolean, or some kind of enum)

Direct Known Subclasses:

BinaryDecider, MulticlassDecider, NumericDecider

public abstract class Decider<A extends Enum<A>,C>
extends Object

A machine-learned decision function, backed by Weka.

Field Summary

Fields

Modifier and Type	Field and Description
protected TypedAttribute[]	_attributes
protected TypedAttribute	_classAttribute
protected weka.classifiers.Classifier	_classifier
protected weka.core.Instances	_datasetHeader

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	Decider(String name, TypedAttribute[] attributes, TypedAttribute classAttribute, Vector<weka.filters.Filter> filters)

Method Summary

Methods

Modifier and Type	Method and Description
protected weka.core.Instances	createDatasetHeader()
Dataset<A,C>	createNewDataset() Returns a new (empty) dataset that can (when filled) be used to train or test this decider.
abstract C	getDecision(weka.core.Instance instance) Makes a decision for the given instance.
abstract HashMap<C,Double>	getDecisionDistribution(weka.core.Instance instance) Returns a HashMap associating possible decision outputs with a confidence value in that decision.
abstract InstanceBuilder<A,C>	getInstanceBuilder() Returns a builder for constructing an instance that this decider can make a decision for.
String	getName() Returns the name of the decision function
protected double	getRawClassification(weka.core.Instance instance)
protected double[]	getRawDistributionForInstance(weka.core.Instance instance)
boolean	isReady() You must either load(java.io.File) or #build before you can call getDecision(Instance) or getDecisionDistribution(Instance) This method checks that this has been done.
void	load(File file) Loads a prebuilt decider from file
void	save(File file) Serialises the classifier and saves it to the given file.
void	test(Dataset<A,C> data)
protected void	throwIfCannotClassify(weka.core.Instance instance)
void	train(weka.classifiers.Classifier classifier, Dataset<A,C> trainingData) Trains the decider using the given classifier and training data.

Methods inherited from class java.lang.Object

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

Field Detail

_attributes

protected TypedAttribute[] _attributes

_classAttribute

protected TypedAttribute _classAttribute

_datasetHeader

protected weka.core.Instances _datasetHeader

_classifier

protected weka.classifiers.Classifier _classifier

Constructor Detail

Decider

protected Decider(String name,
       TypedAttribute[] attributes,
       TypedAttribute classAttribute,
       Vector<weka.filters.Filter> filters)
           throws Exception

Throws:

Exception

Method Detail

getName

public String getName()

Returns the name of the decision function

Returns:

the name of the decision function

isReady

public boolean isReady()

You must either load(java.io.File) or #build before you can call getDecision(Instance) or getDecisionDistribution(Instance) This method checks that this has been done.

Returns:

true if this is ready to make decisions.

getDecision

public abstract C getDecision(weka.core.Instance instance)
                       throws Exception

Makes a decision for the given instance.

This function is not available until isReady returns true.

You should typically use an instance builder to generate the instance that is passed into this method.

Parameters:

instance - a set of attribute values

Returns:

the decision

Throws:

Exception

getDecisionDistribution

public abstract HashMap<C,Double> getDecisionDistribution(weka.core.Instance instance)
                                                   throws Exception

Returns a HashMap associating possible decision outputs with a confidence value in that decision. This information can be used, for example, to identify cases when the result of getDecision(Instance) is clear-cut, or a close tie with other possible decisions.

The method by which these values are calculated and their usefulness depends on the classifier.

This function is not available until isReady returns true.

You should typically use an instance builder to generate the instance that is passed into this method.

Parameters:

instance -

Returns:

A HashMap associating possible decisions with confidence values.

Throws:

Exception

getRawDistributionForInstance

protected double[] getRawDistributionForInstance(weka.core.Instance instance)
                                          throws Exception

Throws:

Exception

getRawClassification

protected double getRawClassification(weka.core.Instance instance)
                               throws Exception

Throws:

Exception

throwIfCannotClassify

protected void throwIfCannotClassify(weka.core.Instance instance)
                              throws Exception

Throws:

Exception

createNewDataset

public Dataset<A,C> createNewDataset()

Returns a new (empty) dataset that can (when filled) be used to train or test this decider.

Returns:

a new (empty) dataset for this decider.

train

public void train(weka.classifiers.Classifier classifier,
         Dataset<A,C> trainingData)
           throws Exception

Trains the decider using the given classifier and training data.

Throws:

Exception

test

public void test(Dataset<A,C> data)

load

public void load(File file)
          throws Exception

Loads a prebuilt decider from file

Parameters:

file -

Throws:

Exception

save

public void save(File file)
          throws IOException

Serialises the classifier and saves it to the given file.

Parameters:

file - the file to save the classifier to

Throws:

IOException - if the file cannot be written to

createDatasetHeader

protected weka.core.Instances createDatasetHeader()

getInstanceBuilder

public abstract InstanceBuilder<A,C> getInstanceBuilder()

Returns a builder for constructing an instance that this decider can make a decision for.

Returns:

a builder for constructing an instance for this decider can make a decision for.