financial-anomaly-detection

AI Anomaly - quick start

1. Put historical transactions CSV at data/transactions.csv. Required fields: transaction_id, amount, merchant_id, customer_id Optional: is_anomaly (0/1) for evaluation.

2. Train model locally: $ cd src $ python train.py --csv ../data/transactions.csv

   This writes models/iforest_model_v1.pkl

3. Run service with Docker Compose: $ docker-compose up --build

   FastAPI will be available at http://localhost:8000 Health: GET http://localhost:8000/health Score: POST http://localhost:8000/score

4. Import n8n-workflow.json into n8n. Update HTTP Request node URL if needed.

5. Production notes:

   • Build CI to produce model artifact and push to artifact storage.
   • Deploy scoring service behind HTTPS and an API gateway.
   • Implement auth on scoring endpoint.