Loan Census and Outcome Analysis
Explore an accepted-loan census and compare historical paid-versus-charged-off models.
Accepted-loan census
The source contains accepted loans. It has no rejected applications, so it cannot measure approval rates or identify approval causes.
| Recorded status | Loans |
|---|
Resolved loans by grade
The denominator includes only Fully Paid and Charged Off loans; current and other unresolved statuses are excluded.
| Grade | Resolved loans | Charged off | Charged-off share |
|---|
Historical outcome models
Target: Fully Paid versus Charged Off. Inputs are restricted to stated origination-time loan and financial fields. Post-payment fields, last FICO scores, IDs, geography, and census demographic percentages are excluded. Preprocessing is fitted on training data.
| Model | Locked threshold | Test ROC AUC | Test charged-off F1 | Test precision | Test recall |
|---|
The slider displays validation F1 only. Reported test results use the threshold already selected on validation data. No test threshold is tuned here.
Study scope
The census arrives already cleaned; earlier source transformations and the outcome snapshot date are unavailable. Restricting the target to resolved loans introduces maturity and selection effects, especially in recent cohorts. These measurements describe historical resolved outcomes, not an approval model, a prospective default benchmark, or a demonstrated collection strategy.
No individual loan records or borrower application form are displayed.