Livoa LogoLivoa
ENTROPY ML Ops & Machine Learning Model for Real-time Fraud Detection
Unified Database Hub
Redis (Cache)


• Active User Profiles,

• Recent Transactions.

PostgreSQL (RDS)


• Admin Audit Trails

• User Feedback

DocumentDB (MongoDB)


• Raw Transaction

• Event Logs

RoculmentDB (MongoDB) Device ID


• Raw Transaction Event Logs,

• Device ID

HTTPS/REST: Real-time Transaction Data.
Feature Engineering Layer
Temporal Aggregator
Behavioral Encoder
Multi-Layer Detection

Tier 1: XGBoost

Tier 2: LSTM / GNN / Transformer

Explainability Service


SHAP Engine      LIME Module

Feedback Loop


⚡ Live Fraud Feed (WebSocket)

📊 Explainability Panel (SHAP / LIME Visualizations)

✋ Manual Override Interface (Whiteist / Blacklist)

🔔 Alert Configuration / Push

🗂 Feedback & Labeling Gateway

Action: Admin Feedback

Feature Store


Feature Engineering Layer

Feature Store


SS / Parquet

Continuous Learning Pipeline (Retraining Trigger)
SageMaker Pipelines
Continuous Learning Pipeline (Retraining Trigger)
Drift Detection
Monitoring
Drift Detection

Entropy

by slantie

0
0 uses