Utilização de machine learning para identificação de clientes aderentes ao open finance no setor bancário

• Ano de defesa: 2024
• Autor(a) principal: Holanda, Felipe Lima de
• Orientador(a): Marschner, Paulo Fernando
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Católica de Brasília
• Programa de Pós-Graduação: Programa Stricto Sensu em Governança, Tecnologia e Inovação
• Departamento: Escola de Educação, Tecnologia e Comunicação
• País: Brasil
• Palavras-chave em Português: Aprendizado de máquina; Instituição financeira
• Palavras-chave em Inglês: Machine learning; Open finance; Financial institution; Marketing
• Área do conhecimento CNPq: CNPQ::CIENCIAS SOCIAIS APLICADAS::CIENCIA DA INFORMACAO
• Link de acesso: https://bdtd.ucb.br:8443/jspui/handle/tede/3614

In the current scenario of technological transformations and growing competition in the financial sector, financial institutions are looking for innovative ways to stand out and offer differentiated value to customers. In view of this, the aim of this study is to carry out an analysis to identify the profile of customers who are most adherent to Open Finance in a financial institution. It also aims to predict which customers have a high propensity to share their data, in order to strategically target marketing actions and optimize product and service offers according to each customer's needs, improving results. This will make it possible to optimize segmentation campaigns, reduce acquisition costs and customer loss, resulting in a better allocation of marketing resources. The sample used consists of a real database with 29 attributes and 200,000 observations chosen at random, with 100,000 for those who consented to share their data and the rest for those who did not share their data during the period from May 2023 to April 2024. A horserace of machine learning algorithms was used to identify the classifier that best suited the data distribution, thus providing a more reliable estimate. As a result, it was observed that XGBoost provides a better response considering all the metrics used - Accuracy, F1-Score, AUC (Area Under the Curve), Precision and ROC (Receiver Operating Characteristic).