Knime RF vs Weka RF

I would like to get the information about the algorithm/performance of Random Forest node available under 

Analytics>Mining>Decisio Tree Ensemble>Random Forest>Classification> Random Forest Learner


Analytics>Weka>Classification Algorithms> trees>Random Forest


Are both same?

Are their performance same on same dataset?




Hello Arvind,

they are definitely not the same. The Weka Random Forest is part of the Weka integration in the KNIME AP.

As far as I can tell, the Weka Random Forest implementation closely follows Leo Breiman's original Random Forest implementation very closely. In our own implementation we use a number of modifications that have been proposed in recent years. Most of those affect the tree building, which of course has a lot of impact on the whole forest. An example of such a modification is the calculation of binary nominal splits, for which we use techniques that allow a more efficient calculation.

When it comes to performance I believe it is save to say that the KNIME AP Random Forest is in most cases faster (see this e.g. this thread and can at least compete in terms of accuracy/roc (or what kind of metric You are using).

However, Your results will always depend on Your data and trying out both nodes might be worth a shot and is super easy to accomplish in the KNIME AP ;-)

If You would like more details on the implementation feel free to ask.