Model Governance & Version Control
Compare model versions and track performance changes
Compare Model Versions
PML
$22.50M
-4.1%
vs v2.3.0
TVaR (99%)
$18.50M
-3.6%
vs v2.3.0
VaR (99%)
$15.80M
-2.5%
vs v2.3.0
EAL
$2.35M
-1.3%
vs v2.3.0
Model v2.3.1 → PML decreased 4.0% from v2.3.0
Key improvements: Enhanced frequency calibration and updated severity distributions for ransomware scenarios.
Risk Metrics Comparison
Model Performance History
Version History
| Version | Date | Status | EAL | VaR (99%) | TVaR (99%) | PML | Performance |
|---|---|---|---|---|---|---|---|
v2.3.1 | 2024-01-15 | Production | $2.35M | $15.80M | $18.50M | $22.50M | AUC: 0.89 Gini: 0.78 |
v2.3.0 | 2024-01-08 | Archived | $2.38M | $16.20M | $19.20M | $23.45M | AUC: 0.87 Gini: 0.76 |
v2.2.5 | 2023-12-20 | Archived | $2.42M | $16.80M | $19.80M | $24.20M | AUC: 0.86 Gini: 0.74 |
Model Card - v2.3.1
Training Data
Historical loss data (2020-2024)
OSINT indicators (n=1.2M)
Training samples: 15,234
Model Architecture
Frequency: Negative Binomial GLM
Severity: Lognormal mixtures
Copula: Gaussian (correlation)
Limitations
Trained on historical data only
May not capture emerging threats
Requires quarterly recalibration