Global Feature Importance node

[mdibiase2]

Anyone who explain to me how global importance feature scores using RF surrogate model are computed ?

Thanks a lot

[thor_landstrom]

Hello @mdibiase2,

I would take a look at this component description:

Global Feature Importance – knime

Specifically the quote below,

A3) Surrogate Random Forest Model:
Random Forest is trained with optimized parameters “Tree Depth”, “Number of models” and “Minimum child node size”. Feature importance is calculated by counting how many times it has been selected for a split and at which rank (level) among all available features (candidates) in the trees of the random forest.

Hope this helps,
TL

[mdibiase2]

Thank you for responding!

Anyway, i already read that part of documentation, but i wondered if was possible to go more in depth about it.
For instance, what are ranks about ? The tree depth ?
And in particular, is there any example of math computing of the score ?

[k10shetty1]

Hi @mdibiase2,

In context to Random Forest, tree depth is the maximum number of levels a tree can have from the root node to the farthest leaf node. It determines how many splits can be made from the root to a leaf.

In the context of feature importance in a Random Forest, “ranks” refer to the level or depth at which a feature is used to split the data. The root node (the top of the tree) is considered the highest rank (rank 0), the next level (first set of child nodes) is rank 1, and so on.

To understand how the calculations are made, you may explore the component in detail by examining the nodes and the underlying logic used within it.

Best,
Keerthan