Business Context

LendWise, a digital consumer lending platform processing roughly 120K personal loan applications per month, wants a simple and explainable model to predict whether an approved applicant will default within 12 months. The credit risk team needs a baseline model they can review with compliance and deploy quickly.

Dataset

You are given a historical application dataset with one row per funded loan.

• Size: 240K funded loans, 32 input features
• Target: default_12m — whether the borrower defaulted within 12 months of origination
• Class balance: 14.7% default, 85.3% non-default
• Missing data: 12% missing in employment length, 7% in annual income verification fields, <2% elsewhere

Success Criteria

A good solution should achieve strong ranking performance while remaining interpretable enough for model risk review. Target at least ROC-AUC >= 0.78, PR-AUC >= 0.38, and recall >= 0.60 at an operating point where precision >= 0.35.

Constraints

• Prefer a model that is easy to explain to compliance and credit policy teams
• Batch scoring must finish in under 10 minutes for 120K applications
• The training pipeline should be reproducible and easy to retrain monthly
• Avoid target leakage from post-origination variables

Deliverables

1. Build a binary classification model to predict 12-month default risk.
2. Explain why the selected algorithm is appropriate versus more complex alternatives.
3. Define preprocessing for missing values, categorical variables, and numerical scaling.
4. Evaluate the model using threshold-free and threshold-based metrics.
5. Provide feature importance or coefficient interpretation and recommend a decision threshold.