Logistic regression credit default probability (PD) scoring on a synthetic loan portfolio. Computes AUC, KS statistic, and Gini coefficient as model performance metrics; produces a PD distribution; compares IRB (F-IRB and A-IRB) versus Standardised Approach (SA) capital requirements. Chains from ART-05 (EU AI Act Credit-Scoring Conformity) — the conformity check validates the model governance before scoring. Feeds SIM-03 (Basel RWA Scenario Modeler) with PD/LGD inputs for IRB capital computation. All computation client-side — no loan data transmitted.
score_credit_default_riskcredit_assessmentBasel III F-IRB / A-IRBEU AI Act Art. 11BCBS d424EBA GL/2017/16
Educational / Pre-validation only. Synthetic loan portfolio generated client-side using a seeded PRNG — no real borrower data processed. Logistic regression weights are illustrative, not calibrated to any real loan book. IRB capital figures are for scenario analysis ahead of Basel 3.1 (UK PRA PS1/26 January 2027). Chains from ART-05 to confirm EU AI Act model-governance compliance before scoring.
Preset 1
Retail Mortgage
500 loans · LTV-heavy · 2.5% default rate
Preset 2
SME Lending
300 loans · sector mix · 6% default rate
Preset 3
Corporate Book
200 loans · large exposures · 1.5% default rate
Chain Intake — paste AP2 JSON from ART-05 (EU AI Act Credit-Scoring Conformity)
Credit Scoring Results
—
AUC-ROC
Model discrimination
—
KS Statistic
Max default/non-default separation
—
Gini Coefficient
2×AUC − 1
—
Portfolio PD
Weighted avg PD
—
Expected Loss
PD × LGD × EAD
—
IRB RWA
vs SA RWA
PD Distribution (log-score histogram)
ROC Curve (AUC)
KS Plot — Cumulative Default vs Non-Default
RWA Comparison: IRB vs Standardised Approach
Top 10 Highest-PD Loans (Synthetic)
#
PD Score
LTV / DTI
Credit Score
Sector
EL
IRB RWA
—
—
AP2 Artifact Export — execution_hash anchors this computation