Numerical Random forest model (Weka) & Attribute Selection

I'm quite new to Knime and am having a trouble optimisintg my (numerical) Random Forest model. I have created a model usgin the Random Forest Learner/Predictor nodes and it works fine, but I need to perform an attribute selection/optimisation, as I have mroe then 700 descriptors (attributes). I have heard that the Weka Attribute Selected Classifier and started playing with it and reading some tutorials for it, but it doesn't work for me... I wanted to run the Random Forest as a classifier and get the following error:

ERROR AttributeSelectedClassifier (3.7) 0:2977 Execute failed: Unsupported attribute type exception in Weka: weka.classifiers.trees.RandomTree: Cannot handle numeric class!

I also get the same error when I tried using Weka's Random Forest (3.7) node. Im using Knime 3.2.1 and Weka 3.7. Is there any way to fix this? I was reading the forum around Weka nodes and saw some people had similar problems but I didnt' find any way to fix it.

If not, are there other nodes that would allow for attribute selection/optimisation? I have seen there ways to do that in R, but unfortunately I'm not skilled in coding or R and would prefer to avoid learning that software.

Many thanks for your help and time!

We as of today support only weka 3.7.10

The regression for random forest was introduced in 3.7.12

Hello polucas,

I hope You don't mind that I answer to Your question here such that others can also benefit from our conversation.

As Iris already pointed out, there is currently no easy way to make the Weka random forest node work with numerical values as this functionality is not part of the Weka version that is currently integrated in the KNIME AP.

From Your post I guess that what You want to do is select the most important attribute to train Your model. The good news is that there are plenty of ways how You can achieve that using the KNIME AP but in order to get the best results You will have to figure out and undestand how the different approaches work, and which is the best to take (which is not a bad thing in my mind because You will learn a lot of cool things along the way ;)).

To get You started I would suggest looking up the following thread: .

You can also search the forum for terms like dimensionality reduction, feature/attribute selection/elimination, although most of the threads will give similar answers.

I hope that helps.