Detection of shell companies in financial institutions using dynamic social network

Abstract

Shell companies work in financial interaction with other companies to commit several crimes such as concealing resources of illicit origin (money laundering), tax fraud (tax evasion), corruption, bribery, and drug trafficking, among others. This interaction can be represented by a set of nodes and connections that show the multiple relationships between entities over time. The current article proposes to detect transactions related to shell companies in financial systems, using legal person attributes and incorporating self and group comparisons into dynamic social networks. The months of June 2019, September 2020, and November 2021 are taken as evaluation periods to test the proposed methodology. Our methodology performs better than the traditional rules method, yielding balanced accuracies and true positive rates above 0.9 and 0.85, respectively. The false-positive rate was also lower in the proposed model than in the rule system for most evaluation periods. The latter translates into a reduction in costs by compliance investigations.