Explainability Dashboard
SHAP Summary
Model Performance
Version
2.3.1
AUC: 0.89
Gini: 0.78
Log Loss: 0.42
Training Samples: 15,234
Trained: 2024-01-15
Feature Contributions
NIRS Score
0.42Impact: high
Third-Party Exposure
0.35Impact: high
CVE Count
0.28Impact: medium
Email Spoofability
0.22Impact: medium
Cloud Misconfig
0.18Impact: medium
Dark Web Signals
-0.15Impact: low
DNS/Domain Score
-0.08Impact: low
Natural Language Reasons
Delta Analysis vs. Last Run
NIRS decreased by 3.2 points (from 45.7 to 42.5)
Third-party exposure increased by 8.5%
CVE count decreased by 5 vulnerabilities
Shapley Capital Allocations (Portfolio-Level)
Model Card
Dataset
Historical loss data (2020-2024)
OSINT indicators (n=1.2M)
Limitations
Model trained on historical data
May not capture emerging threats
Fairness Metrics
Demographic Parity: 0.04
Equalized Odds: 0.06