Model Stacking question with pre-computed models

Machine Learning and Modeling
1

I am a beginner in machine learning with R, and I am trying to do model stacking. My models however, correspond to a position in an enzyme and a classification value from the model, which I have already pre-computed, and then, for some positions, I also have whether this position is pathogenic, benign or unknown. For example:

Position    Model1               Model2             Status
1           likely_Pathogenic   likely_Pathogenic  Pathogenic
2           likely_Pathogenic   likely_Benign      Benign
3           likely_Pathogentic  uncertain          Unknown
etc

I would like to do model stacking in order to have a meta-model that would predict if an unknown position is benign or pathogenic. I am using the parsnip and the stacks package but I am not clear how to specify a pre-computed model such as mine.

Any suggestions?

2

Today {stacks} still don't support custom models.

If I understand well your case, I think that you can directly tune penalization and mixture hyperparameters of a "Status ~ Position + Model1 + Model2" multinom_reg model using glmnet with this data if you create a resample of it . Then use the best fit to generate the predictions on test.