Balancing Precision and Recall in Risk Models

Requirements Clarification & Assessment

1. Understanding Business Context:

   • The primary goal of the mortgage bank is to identify potential defaulters accurately to mitigate financial risks.
   • The bank is more concerned about minimizing losses associated with defaults rather than the operational costs of investigating false positives.

2. Model Performance Measures:

   • Precision: Measures the proportion of true defaulters among those predicted as defaulters (TP / (TP + FP)).
   • Recall: Indicates the proportion of actual defaulters that were correctly identified (TP / (TP + FN)).
   • The current model has high recall but low precision.

3. Business Implications:

   • High recall suggests that most defaulters are correctly identified, reducing the risk of unexpected defaults.
   • Low precision indicates a significant number of non-defaulters are incorrectly flagged, leading to unnecessary investigations and potential customer dissatisfaction.

4. Stakeholder Priorities:

   • Prioritize reducing the risk of defaults over minimizing operational costs associated with false positives.
   • Maintain customer trust and satisfaction by minimizing unnecessary interventions.