What are limitations validation set of approach over k-fold cross validation?



I am currently studying about the methods by which I can estimate the test error while building the model on training data.While studying I came across two methods first is validation approach and other is k-fold cross validation approach.

Validation approach- In this we randomly divide the given data set of samples into two parts

  1. training set
  2. validation set

k-fold cross validation- In this we randomly divide the data into K equal-sized parts. We leave out part k, fit the model to the other K - 1 parts (combined), and then obtain predictions for the left-out kth part.

I want to know the limitations of validation approach over k-fold cross validation approach.


Validation approach divides data into two parts training set and test set. So you have less data to train your model and same with test. It makes the model biased. But in K-fold cross validation more data took part in training and testing bcz we remove a small part of data for testing and while iterating we use complete set of data.


Hi @harry

One of the libation is computer intensiveness, you will have to repeat your learning and testing K time. If you take 100 or one left out then this can take a lot of time. Higher k lees bias but …
I take this from “Applied predictive modelling” page 70, The choice of K is usually 5 or 10 but there is no formal rule… A unbiased method may pay a high price in uncertainty… K fold has generally one high variance compare to other method…

Have a good day