Compare classifiers' performance

I want to training some classifiers to compare their performance on test data's prediction. I have to use the cross validation method;i know this is possible in Knime by "X-validation" node.

I've impleted my issue with "Counting loop start"--->"metanode"--->"Loop end" nodes:

- the external loop, at each iteration, chooses a different classifier

- in the metanode I put 4(equals to number of classifiers i want to use) "Cross validation" nodes. In each of them there are "X-validation","X-Aggregator", classifier's learner and prediction nodes.

But in this way, at each iteration, the "X-validation" node generates different training and test data.

So at the end, the classifiers' performance will be compared using different data.

For more precision, I have to to produce the same traininig and test data . How can i use only one "X-validation" node for all classifiers?


I would do it this way (actually I did it this way):

End IF

Hi filips, thanks for your answer but i need to use the cross- validation method that includes X-partitioner and X-aggregator.

In your workflow, to generate training and test data you use the partitioning node. This is correct but is different from cross-validation method.