Is AIFactorValidation() supported as a feature inside another AI Factor?
I'm testing whether an AI Factor's saved validation predictions can be used as a feature in another AI Factor (stacked/meta-model).
P123 accepts the following as a feature:
AIFactorValidation("OM Growth Predictor V1 w transactions","lightgbm fast 2")
The source AI Factor has saved validation predictions from 03/10/2012 through 01/24/2026 every 4 weeks.
However, when loading the second AI Factor, it requests dates that don't correspond to available saved predictions. I've tried matching the source AI Factor's frequency and changing the second model's Dataset Period/start date, but the requested dates still don't consistently align with the saved validation prediction dates.
A more robust solution would be to use Imported Factors. The steps are: 1) download the validation data 2) import it back as an Imported Factor 3) use the imported factor as an AI Factor feature.
We recently made some upgrades to Imported Factors, most crucial is the ability for values to expire (resolve to N/A) after a number of days that you decide. This will make it very robust for the alignment issue, but you can still end up with lots of NAs. You can read about it here: Imported Factor Upgrades (now includes a staleness filter)
The coding is pretty easy. There are several examples to get you started.