keel.GraphInterKeel.datacf.partitionData.PartitioningSchemes

Class DistanceBased_best

java.lang.Object

keel.GraphInterKeel.datacf.partitionData.PartitioningSchemes.DistanceBased_best

public class DistanceBased_best
extends java.lang.Object

This class implements a stratified scheme (equal number of examples of each class in each partition) to partition a dataset

Constructor Summary

Constructors

Constructor and Description
DistanceBased_best(java.lang.String source_file, int np) It reads the training set and creates the partitions

Method Summary

Modifier and Type	Method and Description
void	calculo_previo_hvdm()
void	createPartitionFiles(java.lang.String _carpeta, java.lang.String _ds) It creates the files of each training and test partition
void	deletePartitionFiles() It deletes the files of each training and test partition
double	distancia(int ej1, int ej2) Calculates the HVDM distance between two instances
int	evaluationKNNClass(int nvec, int instancia, int nClases, boolean distance, int[] vecinos, int clase) Computes the k nearest neighbors of a given item belonging to a fixed class.
Instance[]	getInstances() It returns all the original instances
java.util.Vector[]	getPartitions() It returns the indexes of the original instances in all partitions
Instance[]	getTestPartition(int num) It returns the test partition specified
Instance[]	getTrainPartition(int num) It returns the training partition specified

Methods inherited from class java.lang.Object

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

Constructor Detail

DistanceBased_best

public DistanceBased_best(java.lang.String source_file,
                          int np)

It reads the training set and creates the partitions

Method Detail

getTrainPartition

public Instance[] getTrainPartition(int num)

It returns the training partition specified

Parameters:

num - number of the partition

Returns:

the training partition

getTestPartition

public Instance[] getTestPartition(int num)

It returns the test partition specified

Parameters:

num - number of the partition

Returns:

the test partition

getInstances

public Instance[] getInstances()

It returns all the original instances

Parameters:

the - instances

getPartitions

public java.util.Vector[] getPartitions()

It returns the indexes of the original instances in all partitions

Parameters:

the - indexes of the instances in each partition

createPartitionFiles

public void createPartitionFiles(java.lang.String _carpeta,
                                 java.lang.String _ds)

It creates the files of each training and test partition

deletePartitionFiles

public void deletePartitionFiles()

It deletes the files of each training and test partition

evaluationKNNClass

public int evaluationKNNClass(int nvec,
                              int instancia,
                              int nClases,
                              boolean distance,
                              int[] vecinos,
                              int clase)

Computes the k nearest neighbors of a given item belonging to a fixed class. With that neighbors a suggested class for the item is returned.

Parameters:

nvec - Number of nearest neighbors that are going to be searched

conj - Matrix with the data of all the items in the dataset

real - Matrix with the data associated to the real attributes of the dataset

nominal - Matrix with the data associated to the nominal attributes of the dataset

nulos - Matrix with the data associated to the missing values of the dataset

clases - Array with the associated class for each item in the dataset

ejemplo - Array with the data of the specific item in the dataset used as a reference in the nearest neighbor search

ejReal - Array with the data of the real attributes of the specific item in the dataset

ejNominal - Array with the data of the nominal attributes of the specific item in the dataset

ejNulos - Array with the data of the missing values of the specific item in the dataset

nClases - Class of the specific item in the dataset

distance - Kind of distance used in the nearest neighbors computation. If true the distance used is the euclidean, if false the HVMD distance is used

vecinos - Array that will have the nearest neighbours id for the current specific item

clase - Class of the neighbours searched for the item

Returns:

the majority class for all the neighbors of the item

distancia

public double distancia(int ej1,
                        int ej2)

Calculates the HVDM distance between two instances

Parameters:

ej1 - First instance

ej1Real - First instance (Real valued)

ej1Nom - First instance (Nominal valued)

ej1Nul - First instance (Null values)

ej2 - Second instance

ej2Real - First instance (Real valued)

ej2Nom - First instance (Nominal valued)

ej2Nul - First instance (Null values)

Euc - Use euclidean distance instead of HVDM

Returns:

The HVDM distance

calculo_previo_hvdm

public void calculo_previo_hvdm()