I am using the feature importance of logistic regression to rank the most influential features. My problem lies that I do not how to exactly interpret the results of the feature importance. I know that positive and negative values push the classifier towards the positive and negative class respectively. If I get a positive value equal to 2, I do not know what that exactly means. Does the influence logistic or linear and how much is the influence numerically?
Asked
Active
Viewed 31 times
0
-
When you say "feature importance" are you referring to the coefficients or something else? – Jonny Lomond Aug 01 '21 at 13:43
-
i refer to the magnitudes of the coefficients – Elena Stamatelou Aug 01 '21 at 13:50
-
Does this answer your question? Ranking features in logistic regression – Joe Janssen Aug 01 '21 at 18:15
-
https://stats.stackexchange.com/questions/106344/how-to-quantify-the-relative-variable-importance-in-logistic-regression-in-terms may help. – rolando2 Aug 01 '21 at 20:46
-
1This book covers different feature importance methods Interpretable Machine Learning. Model agnostic ones could be helpful here. Magnitude of the coefficients may not be sufficient for reliable ranking. – patagonicus Aug 01 '21 at 22:09