org.android.activityminer.logic.classification
Class KNN

java.lang.Object
  org.android.activityminer.logic.classification.KNN

All Implemented Interfaces:

Classifier

public class KNN
extends Object
implements Classifier

Field Summary

private String[]	attributeNames The attribute names per point index of a trained KNN instance
private boolean	doNormalize The normalization configuration flag.
private KdTree<Long>	kdTree The k-dimensional tree.
private int	neighborCount The count of neighbors.
private Normalizer	normalizer The attribute value normalizer.

Constructor Summary

KNN(int neighborCount, boolean doNormalize)
Constructor

KNN(List<ModelElement> pmmlModelElements, ActivityMinerExtension extension)
Constructor

Method Summary

protected void	addPoint(double[] point, long cId) Does add a data point to the tree.
PredictionResult	classify(Activity activity) Does classify the given sample
private void	fromPMMLModel(NearestNeighborModel nearestNeighborModel) Method to create the classifier model from a PMML nearest neighbor model.
void	fromPMMLModelElements(List<ModelElement> modelElements, ActivityMinerExtension extension) Method to create the classifier model from PMML model elements.
String[]	getAttributeNames() Getter for the known attribute names
String	getDescription(Context context) Getter for a significant description of the configured classifier.
static String	getDescription(Context context, int neighborCount, boolean normalized) Getter for a significant description of a classifier with the given configuration parameters.
KdTree<Long>	getKdTree() Getter for the k-dimensional tree
int	getNeighborCount() Getter for the neighbor count
Normalizer	getNormalizer() Getter for the normalizer
protected void	prepare(String[] attributeNames, int size) Does prepare the classifier.
List<ModelElement>	toPMMLModelElements(Context context, DataDictionary dataDictionary) Getter for the PMML model elements.
void	train(Set<? extends Activity> trainingSet) Does initialize and train the classifier (internal model creation)
private void	trainPMMLModel(List<Row> rows, String[] orderedAttributeCols) Does train the classifier model from PMML data

Methods inherited from class java.lang.Object

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

Field Detail

kdTree

private KdTree<Long> kdTree

The k-dimensional tree.

attributeNames

private String[] attributeNames

The attribute names per point index of a trained KNN instance

neighborCount

private int neighborCount

The count of neighbors.

doNormalize

private boolean doNormalize

The normalization configuration flag.

normalizer

private Normalizer normalizer

The attribute value normalizer.

Constructor Detail

KNN

public KNN(int neighborCount,
           boolean doNormalize)

Constructor

Parameters:

neighborCount - the count of nearest neighbors to use for decision

doNormalize - if true attributes will be normalized to the range of [0,1]

KNN

public KNN(List<ModelElement> pmmlModelElements,
           ActivityMinerExtension extension)

Constructor

Parameters:

pmmlModelElements - the PMML model elements to build from

extension - the activity mine specific extensions

Method Detail

getAttributeNames

public final String[] getAttributeNames()

Getter for the known attribute names

Returns:

the known attribute names

getNeighborCount

public final int getNeighborCount()

Getter for the neighbor count

Returns:

the neighbor count

getKdTree

public final KdTree<Long> getKdTree()

Getter for the k-dimensional tree

Returns:

the k-dimensional tree

getNormalizer

public Normalizer getNormalizer()

Getter for the normalizer

Returns:

the normalizer

train

public final void train(Set<? extends Activity> trainingSet)

Description copied from interface: Classifier

Does initialize and train the classifier (internal model creation)

Specified by:

train in interface Classifier

Parameters:

trainingSet - the training set

prepare

protected final void prepare(String[] attributeNames,
                             int size)

Does prepare the classifier.

Parameters:

attributeNames - the attribute names for the point dimensions

size - the tree size

addPoint

protected final void addPoint(double[] point,
                              long cId)

Does add a data point to the tree.

Parameters:

point - the data point to add

cId - the class identifier

classify

public final PredictionResult classify(Activity activity)

Description copied from interface: Classifier

Does classify the given sample

Specified by:

classify in interface Classifier

Parameters:

activity - the activity to classify

Returns:

the predicted class identifier or -1 if no prediction is made

getDescription

public final String getDescription(Context context)

Description copied from interface: Classifier

Getter for a significant description of the configured classifier.

Specified by:

getDescription in interface Classifier

Parameters:

context - the context

Returns:

the description of this classifier instance

getDescription

public static final String getDescription(Context context,
                                          int neighborCount,
                                          boolean normalized)

Getter for a significant description of a classifier with the given configuration parameters.

Parameters:

context - the context

neighborCount - the neighbor count

normalized - normalization flag

Returns:

the description

toPMMLModelElements

public final List<ModelElement> toPMMLModelElements(Context context,
                                                    DataDictionary dataDictionary)

Description copied from interface: Classifier

Getter for the PMML model elements. The mining schema and the output fields can be ignored, they will be overridden by the calling instance based on the actual configuration.

Specified by:

toPMMLModelElements in interface Classifier

Parameters:

context - the context

dataDictionary - the data dictionary

Returns:

the PMML model

fromPMMLModelElements

public final void fromPMMLModelElements(List<ModelElement> modelElements,
                                        ActivityMinerExtension extension)

Description copied from interface: Classifier

Method to create the classifier model from PMML model elements.

Specified by:

fromPMMLModelElements in interface Classifier

Parameters:

modelElements - the model elements to build model from

extension - the activity mine specific extensions

fromPMMLModel

private final void fromPMMLModel(NearestNeighborModel nearestNeighborModel)

Method to create the classifier model from a PMML nearest neighbor model.

Parameters:

nearestNeighborModel - the knn model

trainPMMLModel

private void trainPMMLModel(List<Row> rows,
                            String[] orderedAttributeCols)

Does train the classifier model from PMML data

Parameters:

rows - the rows with the training data

orderedAttributeCols - the column names in the order of corresponding attribute names