When working with smaller data sets, a e.g. frequency based estimation of class probability may be too optimistic and therefore, the Laplace correction can be used to introduce some uncertainty.
Do the KNIME predictor nodes make use of this correction by default ? If not, would it be conceivable to add such a feature in future ?
EDIT: see my answer here below.
I've performed a quick analysis on the Decision Tree Learner and Predictor. Laplace correction does not appear to be applied.
I would therefore add it as a feature request: a simple tick box in the concerned Learner nodes allowing to apply the said correction.
we (what you definitely know) have quite some predictor nodes. And I don't see that we will incorporate a Laplace correction in all of them in the near future. It is a great idea to have it, but we first need free developer resources for it.
Anyway, I open a feature request for the Decision Tree for now and as soon as it is included I will let you know here.
Thank you a lot for the feedback, Iris !
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.