Published Study no.22 by UIF on machine learning to identify companies having connections with Organized Crime

On June 11, 2024, the Financial Intelligence Unit (FIU) has published Study no.22 which examines the development of a machine learning algorithm designed to identify companies which may have connection with Organized Crime (“OC”) in Italy.
This model utilizes a dataset of 1.804.278 Italian enterprises in the period 2010-2021, which includes financial sheets data, debt exposure to the banking and financial system, employment data, and information on ownership and governance structure.
A sample of 28,570 enterprises at high risk of linking to organized crime, based on public sources and selections from the FIU, was used to train the model.
The model’s ability to correctly recognize non-infiltrated enterprises (so-called Specificity) is 74.2 percent.
The algorithm allows the calculation of a risk indicator for more than 900,000 active companies in Italy, using balance sheet data from 2021.
Many model validation exercises have been carried out using independent data such as STRs and Italian Financial Police data.
The risk indicator has possible applications such as developing territorial or sectoral risk maps and supporting the institutional functions of the FIU and investigative bodies.