split data to train,test and validation

I want to split my data into train and test in a ratio of 70:30,further I want to split my train data into train and validation in a ratio of 60:10. I want 5 folds of such train,test and validation data combination but test data should be same in all 5 folds.

How can I solve this ?

Can any one help me.

Answer 1

[Adam Danz]

If you have the Statistics and Machine Learning toolbox, use cvpartition.

Otherwise, randperm to create indicies that separate the test and training sets.

[Fathima Bareeda]

How can i solve this using randperm

[Adam Danz]

For example, let's say I want to split up vector "data" into 70/30 training/test sets.

data = rand(1,1000); 
trainIdx = randperm(numel(data), round(numel(data)*.7));
testIdx = find(~ismember(1:numel(data), trainIdx));

trainIdx and testIdx are indices of the 'data' vector without overlap. To confirm that,

isequal(sort([trainIdx,testIdx]), 1:numel(data))

ans = logical
   1

The training set will roughly cover 70% while the test set will roughly cover 30%. To confirm that,

[numel(trainIdx)/numel(data),  numel(testIdx)/numel(data)]

ans = 1×2

0.7000 0.3000

●