SAS Miner: Forest Model issues

DarioM — Thu, 14 Jul 2022 16:32:53 GMT

Hi, I am making various models for rare target data. I have made log. Regression, regression, ensemble, and gradient boosting. When I model compare, I find different chosen models for different “selection statistic” metrics.

I am wondering, how do I check my models final variables it selected? How do I find the feature importance chart of any kind?

Also, I am wanting to use random forest model as well, but I get an error when I use 20 samples, apply LARS and partition the data, not sure what the step before the Forest node is (because the old version of miner had a node for this but I am using 14.2 and it does not exist), so I get an error when trying to run the forest. It says “must use at least one input or rejected variable”, and I am not sure how to fix this and get the forest to run. Below is an image of my forest models flow.

Thanks.

Re: SAS Miner: Forest Model issues

GuyTreepwood — Thu, 14 Jul 2022 22:11:42 GMT

Re: feature selections....Did you check the results of the model nodes that you ran? You should be able to see which variables were used/selected for that specific node. Regarding feature importance, I believe the tree-based model nodes have importance stats when you open the results screens.

topic Re: SAS Miner: Forest Model issues in SAS Data Science

SAS Miner: Forest Model issues

Re: SAS Miner: Forest Model issues