keel.Algorithms.Preprocess.Missing_Values.EventCovering

Class EventCovering

java.lang.Object

keel.Algorithms.Preprocess.Missing_Values.EventCovering.EventCovering

public class EventCovering
extends java.lang.Object

Based on the work of Wong et al., a mixed-mode probability model is approximated by a discrete one. First, they discretize the continuous components using a minimum loss of information criterion. Treating a mixed-mode feature n-tuple as a discrete-valued one, the authors propose a new statistical approach for synthesis of knowledge based on cluster analysis. As main advantage, this method does not require neither scale normalization nor ordering of discrete values. By synthesis of the data into statistical knowledge, they refer to the following processes: 1) synthesize and detect from data inherent patterns which indicate statistical interdependency; 2) group the given data into inherent clusters based on these detected interdependency; and 3) interpret the underlying patterns for each clusters identified.

The method of synthesis is based on author's eventcovering approach. With the developed inference method, we are able to estimate the MVs in the data. This method assumes the data is DISCRETIZED (but won't throw any error with continuous data).

Since:

JDK 1.5

Version:

0.5

Author:

Written by Julián Luengo Martín 06/03/2006

Constructor Summary

Constructors

Constructor and Description
EventCovering(java.lang.String fileParam) Creates a new instance of EventCovering

Method Summary

Modifier and Type	Method and Description
protected java.util.Vector	clusterInitation(double[] Px) Initializes the set of clusters using information of the data set
protected double[][]	computeMutualInformation() Estimates the mutual information between the instances in the data set
protected double[]	computePx(java.util.Vector tree) Computes the conjunctive probabilities using the second order probabilities.
protected java.util.Vector	computeTree(double[][] I) Computes the dependece Tree using Dijkstra algorithm
protected double	D(Instance x, java.util.Vector S) Computes the minimum Hamming distance between the instance given and one of the instances in the set given.
protected double	dist(Instance i1, Instance i2) Computes the Hamming distance between 2 instances
protected double	NS(Instance inst, int numCluster, int sizeCluster, double[] R, FreqListPair[] acj_xk, java.util.Vector[] Ekc, double NS_denom) Returns the NS value of a instance given.
void	process() Process the training and test files provided in the parameters file to the constructor.
protected java.util.Vector	refineClusters(java.util.Vector Clusters) This method refines the initial clusters obtained by clusterInitiation()

Methods inherited from class java.lang.Object

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

Constructor Detail

EventCovering

public EventCovering(java.lang.String fileParam)

Creates a new instance of EventCovering

Parameters:

fileParam - The path to the configuration file with all the parameters in KEEL format

Method Detail

dist

protected double dist(Instance i1,
                      Instance i2)

Computes the Hamming distance between 2 instances

Parameters:

i1 - First Instance

i2 - Second instance

Returns:

The Hamming distance

computeMutualInformation

protected double[][] computeMutualInformation()

Estimates the mutual information between the instances in the data set

Returns:

The mutual information for all possible combinations of 2 instances

computeTree

protected java.util.Vector computeTree(double[][] I)

Computes the dependece Tree using Dijkstra algorithm

Parameters:

I - The paired-mutual information of this data set

Returns:

computePx

protected double[] computePx(java.util.Vector tree)

Computes the conjunctive probabilities using the second order probabilities.

Parameters:

tree - The dependence tree of this data set

Returns:

An array of probabilites for each instance (in the same order)

clusterInitation

protected java.util.Vector clusterInitation(double[] Px)

Initializes the set of clusters using information of the data set

Parameters:

Px - The second order probablity estimation

Returns:

a initinal set of clusters

refineClusters

protected java.util.Vector refineClusters(java.util.Vector Clusters)

This method refines the initial clusters obtained by clusterInitiation()

Parameters:

Clusters - The set of clusters to be refined

Returns:

A refined set of clusters

NS

protected double NS(Instance inst,
                    int numCluster,
                    int sizeCluster,
                    double[] R,
                    FreqListPair[] acj_xk,
                    java.util.Vector[] Ekc,
                    double NS_denom)

Returns the NS value of a instance given.

Parameters:

inst - instance given.

numCluster - number of cluster considered.

sizeCluster - size of the cluster.

R - Double array R parameter.

acj_xk - Frequencies list pair.

Ekc - Ekc matrix.

NS_denom - NS fraction to divide the NS value.

Returns:

the NS value of a instance given.

D

protected double D(Instance x,
                   java.util.Vector S)

Computes the minimum Hamming distance between the instance given and one of the instances in the set given.

Parameters:

x - instance given.

S - instances set given.

Returns:

the minimum Hamming distance between the instance given and one of the instances in the set given.

process

public void process()

Process the training and test files provided in the parameters file to the constructor.