Enhancing Anti-Money Laundering: Development of a Synthetic Transaction Monitoring Dataset.

Abstract

Money laundering remains a continuous global problem, necessitating the development of new enhanced transaction monitoring methods. Current anti-money laundering (AML) procedures within the industry are inefficient, and access to transaction monitoring data is limited due to legal and privacy constraints, with available data lacking true labels and diversity. This study presents a new AML transaction generator and uses it to create a dataset called SAML-D. The SAML-D dataset contains 12 features and 28 typologies, expanding beyond the existing datasets by incorporating a wider range of typologies, geographic locations, high-risk countries, and high-risk payment types. The typologies are created based on existing datasets, the literature, and semi-structured interviews with AML specialists. Additionally, machine learning experiments are conducted to present the applicability of the dataset within the field of AML and results are compared to an existing dataset. The primary purpose of the generator and dataset is to provide researchers with an additional resource to evaluate their models and facilitate comparative analysis of their results, potentially assisting the development of more advanced and capable transaction monitoring methods.