classperformance Properties

Classifier performance information

To view the performance-related information of a classifier, create a classperformance object by using the classperf function. Use dot notation to access the object properties, such as CorrectRate, ErrorRate, Sensitivity, and Specificity.

Name and Description

Label — Name of classifier object

'' (default) | character vector

Name of the classifier object, specified as a character vector. Use dot notation to set this property.

Example: 'cp_kfold'

Data Types: char

Description — Description of object

'' (default) | character vector

Description of the object, specified as a character vector. Use dot notation to set this property.

Example: 'performance_data_kfold'

Data Types: char

True Labels and Indices

ClassLabels — Unique set of true labels

Read-only: vector of positive integers | cell array of character vectors

This property is read-only.

Unique set of true labels from groundTruth, specified as a vector of positive integers or cell array of character vectors. This property is equivalent to the output when you run unique(groundTruth).

Example: {'ovarian','liver','normal'}

Data Types: double | cell

GroundTruth — True labels for all observations

Read-only: vector of positive integers | cell array of character vectors

This property is read-only.

True labels for all observations in your data set, specified as a vector of positive integers or cell array of character vectors.

Example: {'ovarian','liver','normal','ovarian','ovarian','liver'}

Data Types: double | cell

NumberOfObservations — Number of observations

Read-only: positive integer

This property is read-only.

Number of observations in your data set, specified as a positive integer.

Example: 200

Data Types: double

ControlClasses — Indices to control classes from true labels

vector of positive integers

Indices to the control classes from the true labels (ClassLabels), specified as a vector of positive integers. This property indicates the control (or negative) classes in the diagnostic test. By default, ControlClasses contains all classes other than the first class returned by grp2idx(groundTruth).

You can set this property by using dot notation or the 'Negative' name-value pair argument with the classperf function.

Example: [3]

Data Types: double

TargetClasses — Indices to target classes from true labels

vector of positive integers

Indices to the target classes from the true labels (ClassLabels), specified as a vector of positive integers. This property indicates the target (or positive) classes in the diagnostic test. By default, TargetClasses contains the first class returned by grp2idx(groundTruth).

You can set this property by using dot notation or the 'Positive' name-value pair argument with the classperf function.

Example: [1 2]

Data Types: double

Sample and Error Distributions

SampleDistribution — Number of evaluations for each sample

Read-only: numeric vector

This property is read-only.

Number of evaluations for each sample during the validation, specified as a numeric vector. For example, if you use resubstitution, SampleDistribution is a vector of ones and ValidationCounter = 1. If you have a 10-fold cross-validation, SampleDistribution is also a vector of ones, but ValidationCounter = 10.

SampleDistribution is useful when performing Monte Carlo partitions of the test sets, and it can help determine if each sample is tested an equal number of times.

Example: [0 0 2 0]

Data Types: double

ErrorDistribution — Frequency of misclassification of each sample

Read-only: numeric vector

This property is read-only.

Frequency of misclassification of each sample, specified as a numeric vector.

Example: [0 0 1 0]

Data Types: double

SampleDistributionByClass — Frequency of true classes during validation

Read-only: numeric vector

This property is read-only.

Frequency of the true classes during the validation, specified as a numeric vector.

Example: [10 10 0]

Data Types: double

ErrorDistributionByClass — Frequency of errors for each class

Read-only: numeric vector

This property is read-only.

Frequency of errors for each class during the validation, specified as a numeric vector.

Example: [0 0 0]

Data Types: double

Performance Statistics

ValidationCounter — Number of validations

Read-only: positive integer

This property is read-only.

Number of validations, specified as a positive integer.

Example: 10

Data Types: double

CountingMatrix — Classification confusion matrix

Read-only: numeric array

This property is read-only.

Classification confusion matrix, specified as a numeric array. The order of the rows and columns in the matrix is the same as in grp2idx(groundTruth). Columns represent the true classes, and rows represent the classifier prediction. The last row in CountingMatrix is reserved for counting inconclusive results.

Example: [10 0 0;0 10 0; 0 0 0; 0 0 0]

Data Types: double

CorrectRate — Correct rate of classifier

Read-only: positive scalar

This property is read-only.

Correct rate of the classifier, specified as a positive scalar. CorrectRate is defined as the number of correctly classified samples divided by the number of classified samples. Inconclusive results are not counted.

Example: 1

Data Types: double

ErrorRate — Error rate of classifier

Read-only: positive scalar

This property is read-only.

Error rate of the classifier, specified as a positive scalar. ErrorRate is defined as the number of incorrectly classified samples divided by the number of classified samples. Inconclusive results are not counted.

Example: 0

Data Types: double

LastCorrectRate — Correct rate of classifier during last run

Read-only: positive scalar

This property is read-only.

Correct rate of the classifier during the last validation run, specified as a positive scalar. In contrast with CorrectRate, LastCorrectRate only applies to the evaluated samples from the most recent validation run of the classifier performance object.

Example: 1

Data Types: double

LastErrorRate — Error rate of classifier during last validation

Read-only: positive scalar

This property is read-only.

Error rate of the classifier during the last validation run, specified as a positive scalar. In contrast with ErrorRate, LastErrorRate only applies to the evaluated samples from the most recent validation run of the classifier performance object.

Example: 0

Data Types: double

InconclusiveRate — Inconclusive rate of classifier

Read-only: positive scalar

This property is read-only.

Inconclusive rate of the classifier, specified as a positive scalar. InconclusiveRate is defined as the number of nonclassified (inconclusive) samples divided by the total number of samples.

Example: 0

Data Types: double

ClassifiedRate — Classified rate of classifier

Read-only: positive scalar

This property is read-only.

Classified rate of the classifier, specified as a positive scalar. ClassifiedRate is defined as the number of classified samples divided by the total number of samples.

Example: 1

Data Types: double

Sensitivity — Sensitivity of classifier

Read-only: positive scalar

This property is read-only.

Sensitivity of the classifier, specified as a positive scalar. Sensitivity is defined as the number of correctly classified positive samples divided by the number of true positive samples.

Inconclusive results that are true positives are counted as errors for computing Sensitivity. In other words, inconclusive results can decrease the diagnostic value of the test.

Example: 1

Data Types: double

Specificity — Specificity of classifier

Read-only: positive scalar

This property is read-only.

Specificity of the classifier, specified as a positive scalar. Specificity is defined as the number of correctly classified negative samples divided by the number of true negative samples.

Inconclusive results that are true negatives are counted as errors for computing Specificity. In other words, inconclusive results can decrease the diagnostic value of the test.

Example: 0.8

Data Types: double

PositivePredictiveValue — Positive predictive value of classifier

Read-only: positive scalar

This property is read-only.

Positive predictive value of the classifier, specified as a positive scalar. PositivePredictiveValue is defined as the number of correctly classified positive samples divided by the number of positive classified samples.

Inconclusive results are classified as negative when computing PositivePredictiveValue.

Example: 1

Data Types: double

NegativePredictiveValue — Negative predictive value of classifier

Read-only: positive scalar

This property is read-only.

Negative predictive value of the classifier, specified as a positive scalar. NegativePredictiveValue is defined as the number of correctly classified negative samples divided by the number of negative classified samples.

Inconclusive results are classified as positive when computing NegativePredictiveValue.

Example: 1

Data Types: double

PositiveLikelihood — Positive likelihood of classifier

Read-only: positive scalar

This property is read-only.

Positive likelihood of the classifier, specified as a positive scalar. PositiveLikelihood is defined as Sensitivity / (1 - Specificity).

Example: 5

Data Types: double

NegativeLikelihood — Negative likelihood of classifier

Read-only: positive scalar

This property is read-only.

Negative likelihood of the classifier, specified as a positive scalar. NegativeLikelihood is defined as (1 - Sensitivity)/Specificity.

Example: 0

Data Types: double

Prevalence — Prevalence of classifier

Read-only: positive scalar

This property is read-only.

Prevalence of the classifier, specified as a positive scalar. Prevalence is defined as the number of true positive samples divided by the total number of samples.

Example: 1

Data Types: double

DiagnosticTable — Diagnostic table

Read-only: 2-by-2 numeric array

This property is read-only.

Diagnostic table, specified as a two-by-two numeric array. The first row indicates the number of samples classified as positive, with the number of true positives in the first column and the number of false positives in the second column. The second row indicates the number of samples classified as negative, with the number of false negatives in the first column and the number of true negatives in the second column.

Correct classifications appear in the diagonal elements and errors appear in the off-diagonal elements. Inconclusive results are considered errors and are counted in the off-diagonal elements. For an example, see Diagnostic Table Example.

Example: [20 0;0 0]

Data Types: double

More About

Diagnostic Table Example

Suppose that a cancer study of 10 patients yields these results.

Patient	Classifier Output	Has Cancer
1	Positive	Yes
2	Positive	Yes
3	Positive	Yes
4	Positive	No
5	Negative	Yes
6	Negative	No
7	Negative	No
8	Negative	No
9	Negative	No
10	Inconclusive	Yes

Using these results, the function computes the DiagnosticTable as follows:

Version History

Introduced before R2006a