💰 Anti-Money Laundering (AML) Analysis

A high-scale machine learning project focused on detecting suspicious financial transactions using advanced AML techniques.

---

📖 Overview

This project delivers a comprehensive analysis of the SAML-D dataset, focusing on strengthening anti-money laundering (AML) strategies through EDA, preprocessing, and predictive modeling.

A high-performance XGBoost Classifier is implemented to identify suspicious financial behavior.

---

🔧 Features

• 📊 Exploratory Data Analysis (EDA): Statistical insights, visualizations, anomaly detection
• 🧹 Data Preprocessing: Feature engineering, encoding, normalization, skewness correction
• 🤖 Machine Learning Model: XGBoost classifier for fraud detection
• 📏 Evaluation Metrics: ROC-AUC, confusion matrix, TPR/FPR analysis

---

📊 Dataset

• Name: SAML-D (Synthetic Anti-Money Laundering Dataset)
• Size: ~9.5 million transactions
• Features:
  • Transaction details
  • Payment types
  • Laundering indicators
  • Sender/Receiver locations

---

📂 Project Structure

.
├── 📁 data/                # Dataset files
├── 📁 notebooks/           # EDA & model training notebooks
├── 📁 models/              # Saved models (optional)
├── requirements.txt       # Dependencies
└── README.md              # Documentation

Installation

Clone the repository and install the required dependencies:

git clone https://github.com/your-username/AML-Detection.git
cd AML-Detection
pip install -r requirements.txt

Usage

1. Load the dataset:

   import pandas as pd
   df = pd.read_csv("SAML-D.csv")

2. Run the Jupyter Notebook for analysis and training:

   jupyter notebook AML_Analysis.ipynb

3. Train the XGBoost model:

   from xgboost import XGBClassifier
   model = XGBClassifier()
   model.fit(X_train, y_train)

Results

✅ Validation AUC: 0.827
✅ Test AUC: 0.812
✅ TPR (90% threshold): 0.900
✅ FPR: 0.580

Visualizations

• Transaction Distributions 📊
• Fraud Patterns by Payment Type 🔍
• Suspicious Transaction Heatmaps 🔥

Contributing

Contributions are welcome! Feel free to submit pull requests or report issues.

Built with ❤️ to combat financial fraud using Data Science.