Suspicious activities detection for anti-money laundering using machine learning techniques

Lim, Aun Chir (2025) Suspicious activities detection for anti-money laundering using machine learning techniques. Final Year Project, UTAR.

PDF Download (15Mb)

Abstract

In recent years, money laundering activities have evolved rapidly and become a primary concern for governments and financial institutions worldwide. This type of financial crimes causes negative impacts on the integrity and stability of the global banking sector. Traditional rule-based anti-money laundering (AML) systems are static and unable to effectively detect the novel tactics involved in modern money laundering schemes. To solve money laundering, more effective techniques for detecting suspicious transactions must be developed. Machine learning is able to learn complex relationships within large datasets then identify anomalies that deviate from well-defined patterns. This enables machine learning model to detect those suspicious activities more accurately than traditional approaches. The ultimate goal of this project is to improve the efficiency, accuracy and transparency of anti-money laundering efforts in today’s banking sector. The final product is a web-based system, AMLGuard, which incorporates a machine learning model to detect suspicious transactions related to money laundering. XGBoost is selected as the core detection engine due to its superior performance among five supervised machine learning algorithms tested: Random Forest, Naïve Bayes, Support Vector Machine and Artificial Neural Network. Explainable AI techniques are incorporated to provide clear explanations of the model’s decisions for each transaction. Additionally, AI-powered insights are integrated to offer human investigators natural language explanation and recommendation for enhancing their understanding of model output and decision making. Overall, AMLGuard demonstrates the potential of integrating advance

Actions (login required)