In training machine learning models it is believed that a k-fold cross-validation technique, usually offer better model performance in small dataset. Also, computationally inexpensive compare to other training techniques. However, why do most studies use 10-fold cross-validation?
I think you will find your answer in the following discussion:
https://stats.stackexchange.com/questions/49692/why-do-researchers-use-10-fold-cross-validation-instead-of-testing-on-a-validati
you should use three different k-fold cross-validations (i.e., k = 5, 7, and 10) for comparison purposes.
K-fold Cross-Validation (CV) is used to utilize our data better. The higher value of K leads to a less biased model that large variance might lead to over-fit, whereas the lower value of K is like the train-test split approach. Typically, in some research use of k-folds is K = 5 or 10 (Xiong et, al, 2020). In cross-validation, we do more than one split. We choose from 5 to10 folds. The best performance is the best that they may get 10 folds. Less noise and bias
Isaac Kofi Nti
Usually, the k-fold validation usually affects the performance parameters of the training model. If you want to choose the value of k in your work, here are some links that will help you
https://datascience.stackexchange.com/questions/28158/how-to-calculate-the-fold-number-k-fold-in-cross-validation
https://stats.stackexchange.com/questions/27730/choice-of-k-in-k-fold-cross-validation
Please kindly note that there is always a tradeoff between variance and bias. Your CPU power is also a factor when making such a choice.
All the best.
Let me know if this was helpful
Dear Isaac Kofi Nti ,
No need to be 10 specifically. It is k-folds and this k can be any number of your choice, based on the nature and the size of the dataset you are using. Mainly, the cross-validation aims to efficiently validate the performance of the designed model. It is a statistical procedure used to estimate the classification ability of learning models. This procedure has a single parameter called k that refers to the number of groups to which the dataset will be split.
When a certain value for k is chosen, it may be used in place of k in the reference to the model, such as k=5 becoming 5-fold cross-validation. The general procedure is as follows.
K-fold Cross-Validation (CV) is used to utilize our data better. The higher value of K leads to a less biased model that large variance might lead to over-fit, whereas the lower value of K is like the train-test split approach. Typically, in some research use of k-folds is K = 5 or 10 (Xiong et, al, 2020). In cross-validation, we do more than one split. We choose from 5 to10 folds. The best performance is the best that they may get 10 folds. Less noise and bias
There is no requirement to be 10. Depending on the type and size of the dataset you're dealing with, k might be any number.
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training