How to select feature selection from the Forward Feature Selection meta node

[Ashley_Nguyen]

Hi all,

I am currently working on the Data Mining assignment involving the Ionosphere dataset. I have successfully implemented the Forward Feature Selection Meta Node with a Naive Bayes classifier. However, I would like to clarify the correct procedure for extracting the final list of selected features for the transformation stage.

Specifically, should I rely on the list provided in the “Filtered Table” output port of the Meta Node, or should I manually record the features from the “Feature Selection Filter” results in the statistics view?

Furthermore, I noticed that the model reaches its peak performance at a certain iteration before the accuracy fluctuates. Should I select the subset that leads to the absolute highest accuracy, or the one that offers the best balance between simplicity and performance?

Thank you.

Pic1: Result Table

Pic2: Filtered Table

[k10shetty1]

@Ashley_Nguyen Welcome to the Forum!

Either way is fine. The Feature Selection Filter node shows all evaluated subsets and lets you select the one you want based on score or a threshold. The Filtered Table output will then contain only the columns that match the criteria you selected, which you can use for the transformation stage.

For the subset choice, it’s up to you. You can pick the one with the highest accuracy, or a smaller subset with similar performance if you prefer a simpler model.