This repository contains a Financial Intelligence Unit (FIU) case study built on the PaySim synthetic transactions dataset.
The project demonstrates how to detect structuring just below Anti-Money Laundering (AML) thresholds, identify anomalies using machine learning, and visualize temporal risk windows.

1. Load dataset into SQLite (src/01_load_to_sqlite.py).
2. Feature extraction using SQL queries (sql/).
3. Python analytics with Pandas, NumPy, Scikit-learn, Matplotlib.
4. Visualization of suspicious patterns:
   • Transaction clustering near USD 10K AML thresholds.
   • Isolation Forest anomaly detection.
   • Temporal heatmaps (weekday × hour).
5. Reports and visuals stored in the reports/ directory.

⚠️ Note: large raw datasets are ignored from Git to stay under GitHub’s file limits.
You can regenerate these visuals by running the scripts in src/.

FIU-PaySim/ ├── sql/ # SQL feature extraction queries ├── src/ # Python analytics scripts ├── reports/ # Visuals and structured outputs ├── requirements.txt # Python dependencies ├── requirements.lock.txt └── README.md

• FATF Recommendations (2012)
• FINMA Guidance 05/2023 – Money Laundering Risk Analysis
• Lopez-Rojas, Elmir, Axelsson. PaySim: A Financial Mobile Money Simulator, EMSS 2016
• Dataset: PaySim synthetic transactions (Kaggle “PaySim1”):
  https://www.kaggle.com/datasets/ealaxi/paysim1/data