What's the best way to do cross-validation?

Asked Mar 18 '24 at 11:36

Active Mar 20 '24 at 14:13

Viewed 38 times

-1

Assume that 3-fold CV is used for comparing performance of logistic regression, SVM, and KNN for a 2-class problem. For each fold, the logistic regression provides the following confusion matrices:

Is it common to add confusion matrices together (shown below) in order to determine accuracy?

edited Mar 19 '24 at 05:22

wjktrs

asked Mar 18 '24 at 11:36

Marlon Brando

4

Typically, one would aggregate the quality over blocks. Since your blocks are of the same size, it does not matter whether you aggregate Corrects and Wrongs and then calculate accuracy, or average accuracy per block. However, please think carefully about whether you want to use accuracy at all, which has major issues. – Stephan Kolassa Mar 18 '24 at 12:14

What's the best way to do cross-validation?

0 Answers0