I’m beginner in knime and I’m try to use it for cross-validation. What I’m try to do is depicted in the following picture usin K-fold random selection
Im’ using the follow workflow
I have no idea how to rotate the test folder, tips?
What test folder?
Seems to me that you’d achieve the scheme in your first image by using linear sampling and picking an appropriate number of validations.
Thanks for the replay.
YES I’ve 5 folders of data builded randomly from a dataset of 80 samples (16 data each). I want use 1 folder as test and the other 4 folds as training. Then I want use always the same folder as test and rotate the other. Example 5 folders F1 F2 F3 F4 F5
F5 become test T1
LOOP 1) training F1 F2 F3 F4 test T1
LOOP 2) training F1 F2 F3 F5 test T1
LOOP 3) training F1 F2 F4 F5 test T1
I wonder if is possible do it by X-partitioner instead of make it manually.
I do not really understand your idea. If you use the same folds each time you should get the same results each time
actually, the set of training folds change because serially I replace one fold with test data, only the test data remain the same.
thanks for your answer
This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.