Definition
Measures the degree to which alternative data features in the credit model have shifted in predictive power relative to their importance at training time, detecting feature degradation before it surfaces as default-rate movement.
Formula
AFSI = SUM (w_f x corr_current_f / corr_training_f) / SUM w_f
Proposed formula, pending validation pilots.
At a Glance
Category
AI Credit & Lending Decision Quality
Decision Supported
Should the credit model continue to rely on alternative data features at their current weights, or have shifts degraded specific features enough to warrant retraining?