Deep Neural Networks for Anti Money Laundering Using Explainable Artificial Intelligence

This paper explores the application of machine learning (ML) and explainable AI (XAI) techniques for detecting money laundering in financial transactions. A novel approach is introduced that combines a deep neural network (DNN) with SHapley Additive exPlanations (SHAP) to enhance the transparency and effectiveness of anti-money laundering (AML) systems. The proposed model demonstrates superior performance over benchmark models, achieving high precision (0.994585), recall (0.994500), F1 score (0.994551), and ROC AUC (0.994525) in identifying fraudulent transactions using a synthetic dataset derived from real financial logs. Through a global explainability analysis, key indicators of fraudulent activities, such as high transaction amounts and prolonged transaction durations, are identified. This study contributes to the AML field by improving model accuracy and providing insights into the decision-making processes of complex ML models. Future research will focus on applying local explanations and utilizing larger real world datasets to further enhance model performance and interpretability.