07-12-2017 11:26 AM
Just a general high-level question, fishing for recommended approaches / best practices.
We want to deploy models that take a subset of input, enhance that input by performing a couple lookups, and then run the model using that larger set of inputs. What we're finding is that the "lookup tables" option in Decision Manager isn't really what we're looking for here: we'd like the additional lookups to be defined in DI Studio such that a nightly batch process could easily update the lookup table(s).
The process in question is real-time, so the lookups are not too complex. But we simply don't see how to add custom data retrieval to anything that gets deployed via Decision Manager/Builder.
Any ideas on how to go about that?
07-12-2017 11:39 AM
You may want to provide a brief example of some start data, a few rows of dummy values, some corresponding lookup data and what the result should be after the look up. That may help those that understand DI provide more targetted responses.
I can't suggest as I don't use DI but the wording you are using almost makes me think your definition of "look up" may have some top-spin that I wouldn't use that same term for.
07-12-2017 11:48 AM - edited 07-12-2017 11:50 AM
Thanks for the quick response. A simple example would be, say, deploying a predictive model that uses zip code as an input, but along with that it requires various demographic information that we can retrieve using the zip code. So for example, we want to look up the demographic inputs given the zip code input... all as part of the scoring process.
One easy solution is to tell the calling system to do the lookup in advance, and thus provide both zip and demographic information up front, but unfortunately we are not in a position to make those demands at this time.