perfByField

Performance of classification based on values in an input field

Contents

Syntax

Description

rr=perfByField(dsTrain, dsTest, opt) returns the performance of a given pair of training and test sets based on uniques values in a give field. Note that the values of the given field must be a categorical variable.

Example

		ds=prData('iris');
		separateFieldName='sepal width';
		id=find(strcmp(ds.inputName, separateFieldName));
		m=median(ds.input(id,:));	% Make the first input categorial
		index=find(ds.input(id,:)>m);
		ds.input(id,:)=0;
		ds.input(id,index)=1;
		cvOpt=cvDataGen('defaultOpt');	% Create training & test sets
		cvOpt.foldNum=2;
 		cvOpt.cvDataType='full';
 		cvData=cvDataGen(ds, cvOpt);
		dsTrain=cvData.TS;
		dsTest=cvData.VS;
		opt=perfByField('defaultOpt');
		opt.fieldName=separateFieldName;
		rr=perfByField(dsTrain, dsTest, opt, 1);

1/2: sepal width=0, training rr=93.0233%, test rr=97.5%
2/2: sepal width=1, training rr=100%, test rr=85.7143%

See Also

perfLoo, perfCv4classifier.

Published with MATLAB® R2020b