Scoring code generation coverage<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />The formula depot and score code generation is an exciting new feature in JMP Pro 13.
nascif@jmp has shared this excellent and informative wiki page showing a modeling platform/score code language compatibility matrix. Such a reference may be helpful when positioning/discussing this new feature with your customers.
sww.sas.com/jmpmediawiki/Scoring_code_generation_coverage
A couple of additional points:
- This document is internal use only, and the contents of the document are not for sharing directly with our customers, but it would be helpful for you to familiarize yourself with the scoring coverage should you get specific question related to score code generation like TS track linked below.
- Nascif informed me that audrey.shull and our JMP development summer student did much of the work to put together this resource.
External folks can get access to the compatibility matrix by looking at the appendix of the JMP Discovery Scoring paper:
https://community.jmp.com/docsU/DOC-10124
Re: Scoring code generation coveragePhil,
Per Dan's comment: Here is the track out here on this...basically the scoring formula is not translating to SQL properly, due to the “Logist” JSL function. This has been an issue ever since, but of course has surfaced again due to the JMP Pro 13 Formula Depot ease of covering scoring code to SQL. Mike V. opened a track for the SQL scoring code problem: 7611884396
Re: Scoring code generation coverageThanks, Phil. That is a true statement. If you have customers who have requested this, please let me know and I will get them added to a suggestion defect. This just recently came up from jordanhiller@jmp 's customer, who opened a tech support track.
Re: Scoring code generation coverageHi Dan,
Thanks for this. Really useful.
I've had questions from prospects in the past about deploying propensity (e.g. Churn/No Churn, Default/No Default) models as SQL. It looks like deploying to SQL is not possible for categorical responses. Is this something that will be addressed in future versions?
Thanks,
Phil