|
When performing S fold Cross Validation on k Nearest Neighbour Algorithm, is there any change in optimal k as S increases |
|
Not if k is truly optimal. While cross validation is used as a way to asses model performance, the method itself has a certain amount of bias/error. Increasing the number of folds reduces that error at the expense of computational time. In other words, it improves your understanding of the performance of the model, but doesn't improve (or hurt) the model itself. |
