galafis/ai-financial-fraud-detection
AI-powered fraud detection system for financial transactions. Uses ensemble models, anomaly detection, and real-time scoring to identify fraudulent patterns.
What's novel
AI-powered fraud detection system for financial transactions. Uses ensemble models, anomaly detection, and real-time scoring to identify fraudulent patterns.
Code Analysis
10 files read · 3 roundsA real-time financial fraud detection system that ensembles Random Forest, XGBoost, a Keras neural network, and an autoencoder with a logistic regression meta-model, served via FastAPI with Kafka streaming ingestion and drift monitoring.
Strengths
Genuinely implements four architecturally different ML models with probability calibration and a meta-model, plus vectorized temporal feature engineering with data-leakage prevention via shift(1). The Kafka consumer, monitoring with thread-safe drift detection, and Prometheus-instrumented API show production-oriented thinking.
Weaknesses
The API inference path bypasses the FeatureEngineer class entirely, creating a train/serve skew risk. Test coverage is thin relative to the codebase — feature engineering tests don't exercise the actual FeatureEngineer class, and the backtest is self-described as minimal.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
69
Contributors
1
Files
41
Active weeks
9
Repository
Language
Python
Stars
1
Forks
0
License
MIT