Risk Scoring

Fifteen models.
One calibrated score.

A meta-learner ensemble combines seven detectors into a single probability with full SHAP attribution.

Request a demo See the ensemble
Risk Assessment LIVE
0.89
Calibrated Risk Score
Velocity spike
+0.31
Sanctions proximity
+0.22
Dormant reactivation
+0.14
Graph centrality
-0.06
Scoring Engine

Every score is defensible.

15
ML models in ensemble
SHAP
feature-level explainability
SR 11-7
model governance compliant
Ensemble Scoring

Seven detectors.
One ensemble.

Each detector evaluates a different risk dimension. The meta-learner combines them with Platt calibration.

Detector Ensemble
Statistical
Z-Score
Isolation Forest
Ensemble
Temporal
Time-Series
Velocity
Sliding Window
Graph
Network
Dormant Account
Behavioral
New Velocity
Adaptive
Meta-Learner
GBM + Platt
Calibrated Score 0.89
Explainability

Every score has a reason.

SHAP TreeExplainer decomposes each score into feature-level contributions. Regulators see exactly what drove every decision.

Transaction velocity (24h)+0.31
Sanctions list proximity+0.22
Dormant account reactivation+0.14
Amount relative to baseline+0.09
Counterparty risk profile+0.07

See how your risk scores hold up.

We score a live transfer with full SHAP decomposition during the demo.

Request a demo