如何用较少的变量在SAS中对模型评分

Question

I have the next problem.

I created a random forest in sas with 200 variables.

Now I want to score new observations with the same model but I have only 50 variables (the important ones!).

When I run the score, it crashes and appeared the next message: ERROR: Score input variable var_1 not found in the data set. (And thats ok because I will not going to generate those variables again, they are useless).

The question is, how can I run the model with only the 50 variables?

Thanks,

Answer 1

You may want to rerun the RF model with the selected 50 variables only. You are on the correct path. You just did not conduct the second using the 50 selected variables only. The error means the saved binary score file still holds the first model columns.