Mejora del proceso de debida diligencia y gestión de riesgos PLA-FT en aseguradora mediante ML

Abstract

The project designs a dynamic risk rating system that integrates risk matrix automation and a machine learning model to determine the customer's risk profile. The study is justified by the need to transform manual management into a predictive and standardized process that mitigates money laundering under regulations such as the UAFE. For the development, a hybrid methodology was adopted that combine process management (BPM) for operational redesign and the CRISP-DM approach for the data life cycle. An initial diagnosis was made in the area of compliance, and a prototype was built in line with the Personal Data Protection Law. The results were evident in operational optimization, reducing processing time per customer from 45 to 10 minutes. Likewise, the false positive rate dropped from 85% to 20% due to the accuracy of the model. In conclusion, the transition from a manual model to a predictive model validates the improvement in the efficiency and reliability of the process. The automation of 80% of due diligence strengthens the regulatory position and optimizes the analyst's experience. Keywords: Machine Learning, Automation, Risk Management, Risk Scoring, and Due Diligence