keel.Algorithms.Fuzzy_Rule_Learning.Genetic.ClassifierFuzzySGERD

Class KNN

java.lang.Object

keel.Algorithms.Preprocess.Basic.Metodo

keel.Algorithms.Fuzzy_Rule_Learning.Genetic.ClassifierFuzzySGERD.KNN

public class KNN
extends Metodo

The KNN algorithm tries to find the K nearest instances in the training data, selecting the most present class. Euclidean (L2), Manhattan (L1) and HVDM distances can be used as distance function by the classifier.

Since:

JDK1.4

Version:

1.1

Author:

Written by Salvador García López (University of Granada) 11/07/2004, Modified by Joaquin Derrac (University of Granada) 3/11/2009

Field Summary

Fields

Modifier and Type	Field and Description
protected InstanceSet	referencia Data structures

Fields inherited from class keel.Algorithms.Preprocess.Basic.Metodo

clasesTest, clasesTrain, datosTest, datosTrain, entradas, ficheroSalida, ficheroTest, ficheroTraining, ficheroValidation, nEntradas, nominalDistance, nominalTrain, nulosTrain, realTest, realTrain, relation, salida, stdDev, test, training

Constructor Summary

Constructors

Constructor and Description
KNN(InstanceSet dataset, int value_k) Parameter constructor.

Method Summary

Modifier and Type	Method and Description
void	configuracion(int value_k) Configures the KNN algorithm setting the k value and the outliers thershold k values as the 80% of k.
static int	differentClass(int nvec, int classE, double[][] conj, double[] ejemplo) Returns the number of neighbours of the example given that differ in the class value.
static double	distancia(double[] ej1, double[] ej2) Computes the Euclidean distance between the two examples given.
void	ejecutar(int[] outliers, int[] ExamplesClass) Executes the KNN algorithm on the training dataset to obtain the outliers during the classification.

Methods inherited from class keel.Algorithms.Preprocess.Basic.Metodo

leerConfiguracion, normalizar

Methods inherited from class java.lang.Object

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

Field Detail

referencia

protected InstanceSet referencia

Data structures

Constructor Detail

KNN

public KNN(InstanceSet dataset,
           int value_k)

Parameter constructor. Basic structures will be initialized with the parameters given.

Parameters:

dataset - Training dataset.

value_k - number of nearest neighbours cosidered.

Method Detail

ejecutar

public void ejecutar(int[] outliers,
                     int[] ExamplesClass)

Executes the KNN algorithm on the training dataset to obtain the outliers during the classification. This method will fill the arrays given with the number of instances that belong to each class. If the number of neightbours with a different class of the evaluated is greater that the threshold 'k_dif', the evaluated example is consired an outlier.

Parameters:

outliers - Number of outliers for each classes.

ExamplesClass - Number of instances for each original classes.

configuracion

public void configuracion(int value_k)

Configures the KNN algorithm setting the k value and the outliers thershold k values as the 80% of k.

Parameters:

value_k - number of neighbors considered.

differentClass

public static int differentClass(int nvec,
                                 int classE,
                                 double[][] conj,
                                 double[] ejemplo)

Returns the number of neighbours of the example given that differ in the class value.

Parameters:

nvec - number of neighbours considered.

classE - original class of the example.

conj - Training dataset.

ejemplo - example to evaluate.

Returns:

the number of neighbours of the example given that differ in the class value.

distancia

public static double distancia(double[] ej1,
                               double[] ej2)

Computes the Euclidean distance between the two examples given.

Parameters:

ej1 - first given example.

ej2 - second given example.

Returns:

the Euclidean distance between the two examples given.