com.datumbox.framework.core.machinelearning.clustering

Class Kmeans

java.lang.Object

com.datumbox.framework.core.machinelearning.common.abstracts.AbstractTrainer<MP,TP>

com.datumbox.framework.core.machinelearning.common.abstracts.modelers.AbstractModeler<MP,TP>

com.datumbox.framework.core.machinelearning.common.abstracts.modelers.AbstractClusterer<Kmeans.Cluster,Kmeans.ModelParameters,Kmeans.TrainingParameters>

com.datumbox.framework.core.machinelearning.clustering.Kmeans

All Implemented Interfaces:

Savable, Parallelizable, PredictParallelizable, Trainable<Kmeans.ModelParameters,Kmeans.TrainingParameters>, TrainParallelizable, java.lang.AutoCloseable

public class Kmeans
extends AbstractClusterer<Kmeans.Cluster,Kmeans.ModelParameters,Kmeans.TrainingParameters>
implements PredictParallelizable, TrainParallelizable

This class implements the K-means clustering algorithm supporting different Initialization and distance methods. Several different expansions on the standard algorithm are used to ensure we can handle mixed data (k-representative, k-prototype, etc). References: http://cs.gsu.edu/~wkim/index_files/papers/kprototype.pdf http://ilpubs.stanford.edu:8090/778/1/2006-13.pdf http://www.ima.umn.edu/~iwen/REU/BATS-Means.pdf http://web.cs.swarthmore.edu/~turnbull/Papers/Turnbull_GenreRBF_KDE05.pdf http://thesis.neminis.org/wp-content/plugins/downloads-manager/upload/masterThesis-VR.pdf

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	Kmeans.Cluster
static class	Kmeans.ModelParameters Base class for every ModelParameter class in the framework.
static class	Kmeans.TrainingParameters The AbstractTrainingParameters class stores the parameters that can be changed before training the algorithm.

Nested classes/interfaces inherited from class com.datumbox.framework.core.machinelearning.common.abstracts.modelers.AbstractClusterer

AbstractClusterer.AbstractCluster, AbstractClusterer.AbstractModelParameters<CL extends AbstractClusterer.AbstractCluster>

Nested classes/interfaces inherited from class com.datumbox.framework.core.machinelearning.common.abstracts.AbstractTrainer

AbstractTrainer.AbstractTrainingParameters

Nested classes/interfaces inherited from interface com.datumbox.framework.core.machinelearning.common.interfaces.PredictParallelizable

PredictParallelizable.Prediction

Method Summary

Modifier and Type	Method and Description
PredictParallelizable.Prediction	_predictRecord(Record r) Takes a single record, makes a prediction using the model and returns the result wrapped in the Prediction tuple.
boolean	isParallelized() Getter for the parallelized parameter.
void	setParallelized(boolean parallelized) Setter for the parallelized parameter.

Methods inherited from class com.datumbox.framework.core.machinelearning.common.abstracts.modelers.AbstractClusterer

getClusters

Methods inherited from class com.datumbox.framework.core.machinelearning.common.abstracts.modelers.AbstractModeler

predict

Methods inherited from class com.datumbox.framework.core.machinelearning.common.abstracts.AbstractTrainer

close, delete, fit, getModelParameters, getTrainingParameters, save

Methods inherited from class java.lang.Object

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

Methods inherited from interface com.datumbox.framework.core.machinelearning.common.interfaces.PredictParallelizable

_predictDatasetParallel, _predictDatasetParallel

Method Detail

isParallelized

public boolean isParallelized()

Getter for the parallelized parameter.

Specified by:

isParallelized in interface Parallelizable

Returns:

setParallelized

public void setParallelized(boolean parallelized)

Setter for the parallelized parameter.

Specified by:

setParallelized in interface Parallelizable

_predictRecord

public PredictParallelizable.Prediction _predictRecord(Record r)

Takes a single record, makes a prediction using the model and returns the result wrapped in the Prediction tuple. The method assumes that the KnowledgeBase is be loaded.

Specified by:

_predictRecord in interface PredictParallelizable

Returns: