🕵️‍♂️ Anti Money Laundering Detection using Transaction Data

This project analyzes synthetic banking transactions to detect patterns associated with money laundering. Using a Kaggle dataset, we perform exploratory data analysis and prepare the ground for classification modeling.

📂 Dataset

• Source: Kaggle – Synthetic Transaction Monitoring Dataset
• File: SAML-D.csv (~996MB)
• Key Columns:
  • Time, Date: Transaction timestamp
  • Sender_account, Receiver_account: Pseudonymous account IDs
  • Amount: Monetary value of transaction
  • Payment_type: Type of transaction (e.g., CASH_OUT, TRANSFER)
  • Is_laundering: Target flag (1 = laundering, 0 = legitimate)

🛠️ Tech Stack

• Python 3
• Google Colab
• Pandas, NumPy
• Seaborn, Matplotlib
• Scikit-learn (for metrics and modeling)

📊 Exploratory Data Analysis

We analyze trends in:

• Frequency of laundering by Payment_type
• Amount patterns in laundering vs normal cases
• Group-wise statistical summaries

Sample Insights:

• Certain payment types (e.g., TRANSFER) have higher laundering rates.
• Laundering transactions often involve higher maximum and average amounts than regular ones.

🚀 How to Run (Colab)

1. Open Anti_Money_Laundering.ipynb in Google Colab.
2. Upload your kaggle.json API key.
3. Run the cell to authenticate and download the dataset.
4. Execute the EDA blocks and modify or extend as needed.

Checkout Kaggle of AML:

Kaggle

📈 Planned Extensions

• Train machine learning models (Logistic Regression, XGBoost)
• ROC-AUC and confusion matrix evaluations
• Integrate anomaly detection
• Convert pipeline into real-time dashboard

🧠 Author's Note

This is a synthetic dataset, but it reflects realistic fraud detection challenges. The aim is to develop interpretable workflows for financial surveillance.

📩 Contact

Siddharthan P S
📧 Email: sp8004@nyu.edu
🔗 LinkedIn: Siddharthan P S

📘 Medium Article: Detecting Money Laundering with Python and AML Dataset
Portfolio: Siddharthan P S

✅ Tip

To reduce RAM usage in Colab:

• Use usecols to load selective columns
• Downcast numeric types with pd.to_numeric(..., downcast=)