Threshold For Predictor Importance

5 views (last 30 days)
BP205
BP205 on 23 Feb 2021
Commented: BP205 on 12 Apr 2021
Suppose you have a set of predictors X1, X2, X3, X4,.....Xn, and a Response variable Y. If after computing predictor importance for X1--Xn against Y, how does one determine a threshold that is statistical; not just an arbitrary cut off?

Accepted Answer

Gaurav Garg
Gaurav Garg on 26 Feb 2021
Hi,
To determine threshold in case of decision trees, Information Gain is calculated for each of the nodes. The node with the highest Information Gain is selected as the decision node, and the same process is repeated for the next level nodes as well.
The way Information Gain is calculated can differ from categorical data to numerical data.
For more information, you can look at the following links -
  1. Splitting categorical variables
  2. Viewing decision trees
Please note that Information gain is computed statistically. You can also easily compute Information Gain on your dataset and then compare it with the decision tree returned by the model.
  1 Comment
BP205
BP205 on 12 Apr 2021
Sorry for the late response. Thanks for the answer

Sign in to comment.

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!