My question is a little related with this question Profiler in KNN
We could not define factor only in KNN among of Formula Depot-available features.
Please let me know the reason why.
Thanks in advance
I am not clear about your question. What "features" in Formula Depot that you couldn't use? The earlier post you referenced is why the prediction profiler is not available for KNN when Y is categorical. The reason is KNN doesn't make probability predictions.
I mean the platforms that publish prediction formulas and generate scoring code include: Discriminant, Least Squares Regression, Logistic Regression, Partition, Uplift,K Nearest Neighbors, Naive Bayes, Neural, Latent Class Analysis, Principal Components,
Generalized Regression, PLS, Gaussian Process.
I want to know the reason why KNN couldn't define factors.
It looks to me that the Formula Depot code generation fails to convert some of the JSL functions in the prediction formulas from K Nearest Neighbor. I have contacted JMP Tech Support and will update when I hear back from their investigation.
[Update]
JMP Tech Support confirms that Formula Depot cannot translate the prediction formula from KNN as it is only capable of mapping basic math and control flow operations from JSL. So, it generates placeholder calls instead for the customer to supply their own implementation.