Package deepnetts.eval

Class ConfusionMatrix

java.lang.Object

deepnetts.eval.ConfusionMatrix

public class ConfusionMatrix
extends Object

Confusion matrix contains raw classifier test results. It counts number of true and false predictions with respect to actual/target class of the given examples in test(evaluation) set. Rows correspond to actual/target classes, and columns to predicted Predicted F T Actual/target F TN FP Actual/target T FN TP https://en.wikipedia.org/wiki/Confusion_matrix

Field Summary

Fields

Modifier and Type	Field	Description
static final String	FALSE_NEGATIVE	A label for items falsely classified as negative, which are actually positive.
static final String	FALSE_POSITIVE	A label for items falsely classified as positive, which are actually negative.
static final String	TRUE_NEGATIVE	A label for items classified as negative which are really negative.
static final String	TRUE_POSITIVE	A label for items classified as positive which are really positive.

Constructor Summary

Constructors

Constructor	Description
ConfusionMatrix(String[] classLabels)	Creates a new confusion matrix for specified class labels

Method Summary

Modifier and Type	Method	Description
final int	get(int actualIdx, int predictedIdx)	Returns a value of confusion matrix at specified position.
final int	getClassCount()
final String[]	getClassLabels()
int	getFalseNegative()	How many positive items has been (falsely) classified as negative.
int	getFalseNegative(int clsIdx)
int	getFalsePositive()	Returns number of false positive classifications.
int	getFalsePositive(int clsIdx)
int	getTotalItems()
int	getTrueNegative()
int	getTrueNegative(int clsIdx)
int	getTruePositive()	Return true positive metric for binary classification.
int	getTruePositive(int clsIdx)	Returns true positive metric for specified class idx for multiclass classification.
final void	inc(int actualIdx, int predictedIdx)	Increments matrix value at specified position.
String	toString()

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Details

TRUE_POSITIVE

public static final String TRUE_POSITIVE

A label for items classified as positive which are really positive.

See Also:

• Constant Field Values

TRUE_NEGATIVE

public static final String TRUE_NEGATIVE

A label for items classified as negative which are really negative.

See Also:

• Constant Field Values

FALSE_POSITIVE

public static final String FALSE_POSITIVE

A label for items falsely classified as positive, which are actually negative.

See Also:

• Constant Field Values

FALSE_NEGATIVE

public static final String FALSE_NEGATIVE

A label for items falsely classified as negative, which are actually positive.

See Also:

• Constant Field Values

Constructor Details

ConfusionMatrix

public ConfusionMatrix(String[] classLabels)

Creates a new confusion matrix for specified class labels

Parameters:

classLabels -

Method Details

get

public final int get(int actualIdx,
 int predictedIdx)

Returns a value of confusion matrix at specified position.

Parameters:

actualIdx - target/actual class idx - corresponds to column

predictedIdx - predicted class idx - corresponds to row

Returns:

value of confusion matrix at specified position

inc

public final void inc(int actualIdx,
 int predictedIdx)

Increments matrix value at specified position.

Parameters:

actualIdx - class idx of actual class - corresponds to row

predictedIdx - class idx of predicted class - corresponds to column

getClassCount

public final int getClassCount()

toString

public String toString()

Overrides:

toString in class Object

getTruePositive

public int getTruePositive()

Return true positive metric for binary classification. True positives metric tells us percent of positive examples which are recognized by the classifier as positive. Or in other words percent of correct predictions for the given positive examples.

Returns:

true positive metric for binary classification

getTruePositive

public int getTruePositive(int clsIdx)

Returns true positive metric for specified class idx for multiclass classification. True positive metric tells how many examples are correctly classified as a positive examples of the given class.

Parameters:

clsIdx - Index of class for which true positive value is returned

Returns:

getTrueNegative

public int getTrueNegative()

getTrueNegative

public int getTrueNegative(int clsIdx)

getFalsePositive

public int getFalsePositive()

Returns number of false positive classifications. Items that do not belong to specific class, but they are recognized as they do Only for binary classification

Returns:

getFalsePositive

public int getFalsePositive(int clsIdx)

getFalseNegative

public int getFalseNegative(int clsIdx)

getFalseNegative

public int getFalseNegative()

How many positive items has been (falsely) classified as negative.

Returns:

How many positive items has been (falsely) classified as negative

getClassLabels

public final String[] getClassLabels()

getTotalItems

public int getTotalItems()