Often it is preferable to use the k$-$fold cross$-$validation method, where we partition the data into k subsets, and the one that is left out in each iteration is the $"$

$"$ set.