0

I'm currently working on an algorithm to improve the "feature_importances_" attribute in Random Forest Regressor, since it can potentially be misleading because of the presence "...of high-cardinality features." Nevertheless, I've been reading Classification and Regression Trees (Breiman et al., 1984), Random Forests (Breiman, 2001) and sklearn's documentation for regression trees, but I haven't found any mention of the meaning/interpretation of the "value" attribute of a node within a decision tree. I provided an image for clarification.

Could someone explain what the "value" attribute mean for the root nodeRahmstorf Sealevel/ Temperature model via RF Regressor

I've been researching Random Forest Regressor for my thesis and haven't found any mention of it, besides its interpretation for classifications problems.

Dwight Dinkins
  • 1
  • 2
  • @NickODell Okay, so my understanding is that the value of a node is the average of the target variable that corresponds to a predictor that falls within a range of values? Or is the average of the predictor itself (e.g., the average of of all values x1<= 2)? – Dwight Dinkins Aug 02 '23 at 15:54
  • @NickODell It's the average of the prediction. Sorry, I misread that. Thanks – Dwight Dinkins Aug 02 '23 at 15:56
  • Another way of thinking of it is, "If I stopped at this node in the tree and didn't go further, what would my best guess for the dependent variable be?" – Nick ODell Aug 02 '23 at 15:59
  • @NickODell Ahhh! That makes sense! Thanks again, truly. – Dwight Dinkins Aug 02 '23 at 16:01

0 Answers0