how to predict response using test data after using 'KFold ', 5 in case of SVM

Hi there...after training a model using following code Mdl = fitrsvm(predictortrain,response,'standardize', true, 'kFold', 5) now kindly tell me how can i calculate the response using 'Kfoldpredict' instead of predict and which parameter i have to pass for 'Kfoldpredict'. as i have seperate data for testing kindly let me know if you have any solution.

[Muhammad Kashif]

if you want to use the 'Kfoldpredict' you need to do some step before, i will post an example.

[Saba Yousaf]

kindly post your example please.

Answer 1

[Muhammad Kashif]

once you trained the model. now you want to use 'Kfoldpredict', first you validate your model. e.g;

Mdl = fitcecoc(features_train,labels_train,'Learners',t,'FitPosterior',1,...
    'ClassNames',{'1','2','3','4','5','6','7'},...
    'Verbose',2); 
  CVMdl = crossval(Mdl);                       % cross- validate Mdl
  oosLoss = kfoldLoss(CVMdl);
  label = predict(Mdl,features_test);          % if want to predict
  oofLabel = kfoldPredict(CVMdl);
i hope itwill help you.

[Saba Yousaf]

oky thank you. let me implement this code

[Saba Yousaf]

sir when i run this code CVMdl = crossval(Mdl); getting this error Error using crossval (line 160) At least two inputs are needed. now wht i have to do?

[Muhammad Kashif]

post you error

[Saba Yousaf]

Mdl = fitrsvm(predictortrain,response,'standardize', true, 'kFold', 5) CVMdl = crossval(Mdl); getting error like this "Error using crossval (line 160) At least two inputs are needed"

[Saba Yousaf]

sir come across a point that if in this command

Mdl = fitrsvm(predictortrain,response,'standardize', true,) then using

CVMdl = crossval(Mdl)
by eliminating KFoLd 5 then not getting this error.

if i want to use kFold 5 then how this code will work. because using

 CVMdl = crossval(Mdl); 
this code default Kfold = 10,
that i don't want to use.

[Tanvir Kaisar]

Saba, I am facing the same problem. Did you find the solution to your problem? Can you please share it?

[Mohsin Khan]

You are not setting the right number of parameters;

Try this, will get right output with 5-fold

Mdl = fitrsvm(predictortrain,response,'standardize', true);
CVMdl = crossval(Mdl, 'kfold', 5);

Answer 2

[yi du]

but how to predict the new data?