Classification Learner App - Custom cross validation grouping

I am using the classification learner app - can I can define the grouping for cross validation?

The data set I am using has repeated measures from a number of different subjects. Those repeated measures from within a single subject are not independent and thus should not be randomized to different groups for cross validation. Is it possible to provide such grouping for cross validation with the classification learner app?

Answer 1

[Muskan]

Hi,

To perform a custom cross validation you can use "cvpartition" to create a custom cross-validation partitioning. You can use the 'Leaveout', 'KFold', or 'Holdout' options, and specify the 'Groups' parameter to ensure that data from the same subject is kept together in the same fold.

You can refer to the following documentation of "cvpartition" for more information: https://www.mathworks.com/help/stats/cvpartition.html

I hope this helps!