Is it possible to perfrom cross validation for k = 1 fold (or 1 fold cross validation)

In Matlab 2018. I have used cross-validation method. in Classfier learner app I found that cross validation is given for fold 2- fold 5. I want to perform K=1 fold .

Answer 1

[Anish Walia]

Though it is possible but it does not have any meaning

In K-Fold cross validation method, we divide the data in K folds. Then we use K-1 folds for training it and evaluate/test it on the 1 fold and then repeat it K times with unique fold for evaluating in each iteration.

So K-Fold cross validation on 1 fold would mean dividing data in 1 fold and using 0 (K-1) fold for training, which basically means not training and just testing on that fold

[jaypal singh]

yes your answer is correct. But how i will do this for k=1 in matlba using classfier learner app. Becuase supose how i will test any model witout traing the any model.

[Anish Walia]

Why do you want to do it for k=1?

[jaypal singh]

I have to plot 1 fold to 10fold cross validation curve for ACC, SEN, Specificity. Means 10 fold CV to ACC

like. in attached example.

I already perform 2fold to 10 fold CV. But 1-fold opttion is not there in MATLAB classifier 2018a app.

[Anish Walia]

Okay so i suppose you misunderstood the problem.

The problem just needs 10 fold cross validation.

So divide the data in 10 folds, and basically plot the sensitivity for each fold.

Retain one fold as hold out, train on remaining 9 and plot the sensitivity for the hold out fold.

Then retain another unique fold as hold out, train on remaining 9 and plot sensitivity for this fold as fold 2 in graph

Similiarly repeat this 10 times with taking unique hold out folds and plot as fold 3, fold 4...., fold 10

[jaypal singh]

Hi

Anish

Do you have MATLAB code to perform 10 fold CV and produce ACC , sen spe of each fold.