This project addresses the complex landscape of corporate ownership structures and the identification of their Ultimate Beneficial Owners (UBOs). As businesses expand, their ownership structures become more complex, providing opportunities for concealing illegal activities such as money laundering and tax evasion. Therefore, regulatory authorities are now prioritizing the analysis of these structures in order to identify individuals with absolute decision-making power and control. Existing algorithmic approaches leverage mathematics to identify UBOs. However, these methods demonstrate poor scalability as ownership structures grow, and the convergence of the algorithms is not guaranteed. This project introduces a novel approach to UBO identification, representing ownership structures as graphs and framing the UBO identification task as an inductive link prediction problem. Leveraging publicly accessible ownership data, a graph machine learning model is trained and evaluated for its overall performance and potential to address the existing problems. The result is a machine learning model that provides reliable predictions and, in some cases, mitigates problems of the traditional algorithmic approach. These results underscore the viability of graph-based machine learning as a valuable
method for identifying UBOs within complex ownership structures. A robust machine learning model for UBO identification has significant applications across various industries and scientific domains, strengthening transparency, compliance, and risk mitigation, while combating financial crime, ensuring ethical practices, and supporting responsible resource management.