APPLICATION OF MACHINE LEARNING IN THE SEGMENTATION OF CUSTOMERS ADOPTING OPEN FINANCE IN BRAZIL

Resumo

This study aimed to apply machine learning techniques to identify the profile of customers most likely to adopt Open Finance within a major Brazilian financial institution classified as S1 by the Central Bank of Brazil, meaning it holds assets equal to or greater than 10% of the national GDP or has international relevance. The research employed real, individual-level data, enabling the development of robust predictive models with high practical applicability. Five techniques widely recognized in the literature were evaluated: Random Forest, Support Vector Machine (SVM), Decision Tree, Logistic Regression, and XGBoost. Among the models tested, XGBoost demonstrated the best performance, achieving an AUC of 0.90 and an Accuracy of 0.82. The most relevant predictor of Open Finance adoption was the customer’s digital profile, followed by individuals with income up to R$2,000.00 and/or investments up to R$5,000.00, and customers aged over 64. These findings offer valuable insights for financial institutions and policymakers by highlighting the importance of segmented communication strategies and digital inclusion.