classifier
Class GroupMeans

java.lang.Object
  classifier.GroupMeans

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable

public class GroupMeans
extends java.lang.Object
implements java.io.Serializable, java.lang.Cloneable

Many classification functions need the mean values of each parameter from a group of training data.

This class provides a set of functions which will get the means by group, and also return larger matrixes of the mean values repeated by a grouping index or repeated for a single group which can be used in classifier training and classifier application

Author:

Doug Gillespie

See Also:

Matrix, Serialized Form

Field Summary

private Jama.Matrix	groupMeans
private int[]	groupSize
private static long	serialVersionUID
private int[]	uniqueGroups

Constructor Summary

GroupMeans(Jama.Matrix data, int[] group)
Constructor works out the number of unique groups in group and then constructs a smaller matrix of means of each parameter by group.

Method Summary

GroupMeans	clone()
private int	getGroupIndex(int iD) The group parameter passed to the constructor can contain grouping values which may not start at 0 and may not be continuous.
private int[]	getGroupIndex(int[] iD) As getGroupIndex(iD) but for arrays.
static int	getGroupIndex(int[] uniqueGroups, int iD) Gets the index of a group id (these will often be the same !)
static int[]	getGroupIndex(int[] uniqueGroups, int[] iD) Gets the index of a group id (these will often be the same !)
Jama.Matrix	getGroupMeans()
Jama.Matrix	getGroupMeans(int[] groups) Create a matrix of means, repeated multiple times (e.g.
Jama.Matrix	getGroupMeans(int group, int n) Create a matrix of means for a single group.
int[]	getGroupSize()
int	getNumGroups()
int[]	getUniqueGroups() Get list of unique groups in the data
private void	makeGroupMeans(Jama.Matrix data, int[] group) Takes the raw training data, works out what gorups are present in group and works out the means for each group.
static int[]	unique(int[] groups) Get a unique list of items in the training group

Methods inherited from class java.lang.Object

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

Field Detail

serialVersionUID

private static final long serialVersionUID

See Also:

Constant Field Values

groupMeans

private Jama.Matrix groupMeans

uniqueGroups

private int[] uniqueGroups

groupSize

private int[] groupSize

Constructor Detail

GroupMeans

public GroupMeans(Jama.Matrix data,
                  int[] group)

Constructor works out the number of unique groups in group and then constructs a smaller matrix of means of each parameter by group.

Parameters:

data - matrix of training data

group - group indexes

Method Detail

getGroupMeans

public Jama.Matrix getGroupMeans()

Returns:

matrix of means by training group

getUniqueGroups

public int[] getUniqueGroups()

Get list of unique groups in the data

Returns:

list of unique groups in the data

getGroupSize

public int[] getGroupSize()

Returns:

the number of entries in each group.

getNumGroups

public int getNumGroups()

Returns:

The number of unique groups in the data

getGroupMeans

public Jama.Matrix getGroupMeans(int[] groups)

Create a matrix of means, repeated multiple times (e.g. so it can be subtracted off a set of training data).

Parameters:

groups - grouping to apply. Must contain only the same entries as in the group parameter passed to the constructor.

Returns:

matrix of grouped means.

getGroupMeans

public Jama.Matrix getGroupMeans(int group,
                                 int n)

Create a matrix of means for a single group.

Parameters:

group - index of group

n - number of rows of data to create

Returns:

matrix with n identical rows of group

makeGroupMeans

private void makeGroupMeans(Jama.Matrix data,
                            int[] group)

Takes the raw training data, works out what gorups are present in group and works out the means for each group.

Parameters:

data - training data Matrix

group - grouping variable

getGroupIndex

private int getGroupIndex(int iD)

The group parameter passed to the constructor can contain grouping values which may not start at 0 and may not be continuous.

Therefore, when constructing new means matrixes, it is necessary to use the index of each group parameter to get the right values out of the groupMeans matrix.

Parameters:

iD - group parameter

Returns:

group parameter index in unique list.

getGroupIndex

public static int getGroupIndex(int[] uniqueGroups,
                                int iD)

Gets the index of a group id (these will often be the same !)

Parameters:

uniqueGroups - list of unique groups (sorted)

iD - group ID to find

Returns:

index in group list.

getGroupIndex

private int[] getGroupIndex(int[] iD)

As getGroupIndex(iD) but for arrays.

The group parameter passed to the constructor can contain grouping values which may not start at 0 and may not be continuous.

Therefore, when constructing new means matrixes, it is necessary to use the index of each group parameter to get the right values out of the groupMeans matrix.

Parameters:

iD - group parameter

Returns:

group parameter index in unique list.

getGroupIndex

public static int[] getGroupIndex(int[] uniqueGroups,
                                  int[] iD)

Gets the index of a group id (these will often be the same !)

Parameters:

uniqueGroups - sorted list of unique groups

iD - list of group id's to find indexes of

Returns:

list of indexes.

unique

public static int[] unique(int[] groups)

Get a unique list of items in the training group

e.g. if group was {1, 9, 3, 1, 4, 9, 4} then the return value would be {1, 3, 4, 9}

Parameters:

groups - group data which will contain many instances of few different values.

Returns:

a shorter array of unique values sorted by size.

clone

public GroupMeans clone()

Overrides:

clone in class java.lang.Object