https://github.com/sangramdedge/ai_fraud_detection

End-to-end AI Fraud Detection & Transaction Monitoring project using SQL, Python, ML models, SHAP explainability, and FastAPI integration.
https://github.com/sangramdedge/ai_fraud_detection

ai-ml credit-analytics financial-analysis isolation-forest narrative-ai predictive-modeling python3 random-forest risk-analysis sklearn sql

Last synced: 9 months ago
JSON representation

End-to-end AI Fraud Detection & Transaction Monitoring project using SQL, Python, ML models, SHAP explainability, and FastAPI integration.

• Host: GitHub
• URL: https://github.com/sangramdedge/ai_fraud_detection
• Owner: sangramdedge
• Created: 2025-10-04T01:25:30.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2025-10-04T01:31:18.000Z (9 months ago)
• Last Synced: 2025-10-04T03:32:44.658Z (9 months ago)
• Topics: ai-ml, credit-analytics, financial-analysis, isolation-forest, narrative-ai, predictive-modeling, python3, random-forest, risk-analysis, sklearn, sql
• Language: HTML
• Homepage:
• Size: 853 KB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          








# 🛠 AI Fraud Detection & Transaction Monitoring  

## 📌 Project Overview  

Banks & e-commerce firms lose billions annually due to fraudulent payments.  

This project builds an AI-powered fraud detection system that:  

- Detects suspicious transactions in real time.  

- Minimizes false positives.  

- Provides SHAP-based explanations for compliance officers.  

## 📂 Repository Structure  

- **report/** → Word/PDF project report.  

- **notebooks/** → Exploratory analysis + model building (HTML & Jupyter).  

- **src/** → Python scripts & FastAPI scoring service.  

- **data/** → Sample transactions (demo only).  

## ⚙️ Tech Stack  

- Python (pandas, scikit-learn, XGBoost, SHAP, FastAPI)  

- SQL (PostgreSQL for ingestion & cleaning)  

- Power BI (dashboard design – planned)  

## 📊 Key Results  

- XGBoost ROC-AUC: **0.98**  

- Recall (fraud detection rate): **95%**  

- False Positives reduced to <10% with SHAP interpretability.  

## 🔮 Future Scope  

- Deploy API on AWS Lambda/EC2.  

- Live monitoring dashboards (Power BI/Tableau).  

- Graph Neural Networks for fraud ring detection.