Thanks amy, the trouble is that the test set wants to see a trigger event from a given userid, but that userid might be associated with different companies, countries, emails etc.

Perhaps some clarification would help?

Is the userid an employee of ABSA overseeing some domain and capturing events? Does the userid represent a given product? Here a userid seems quite opague. It does not limit the events to a given country or company even.

The userid represents an individual registered on the platform and performing transactions. Sometimes consultants or bank personnel can take on multiple clients from various companies, in different countries, with different user roles and carry out transactions on their behalf, as such they would have the same userid but different credentials depending on what client they decide to work with at that moment.

Part of the challenge is trying to figure out how to make use(or get rid) of these cases in order to build a more robust model

Thanks amy!!!!!!! That is very clear, I appreciate it.

Note this does break the causality a bit - it is not as if the trigger event originates from some specific country or company e.g. In stead, a trigger is sort of an aggregate event that spans many potential causes.

But ok, if that is the challenge ... let's saddle up our horses ...