Model as a Service
Model as a Service lays out two ways to test risk
This week's page explains a customer-data model recommendation and a separate bureau-score evaluation.
What shipped
The project page now describes two separate studies: use mapped receivables data to recommend an explainable decision tree, or compare licensed, customer-supplied bureau scores against observed outcomes.
It also states the file types, the common-cohort and time-based test approach, and the 30-day file deletion window.