O uso do aprendizado de máquina como ferramenta direcionadora do marketing bancário

• Ano de defesa: 2022
• Autor(a) principal: Soares, Thiago Faria
• Orientador(a): Silva, Thiago Christiano
• 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; Marketing; Mineração de dados; Instituição financeira
• Palavras-chave em Inglês: Machine learning; Data mining; Financial institution
• Área do conhecimento CNPq: CNPQ::CIENCIAS SOCIAIS APLICADAS
• Link de acesso: https://bdtd.ucb.br:8443/jspui/handle/tede/2997

In times of technological disruption and with increased competition, organizations in various industry sectors are forced to look for alternative ways to obtain revenue. Data analysis has established itself as a competitive advantage for this purpose. This work will use popular supervised machine learning techniques, with the objective of directing marketing actions to customers who are more likely to accept the offer to contract the credit card product, based on a set of labeled examples data available for training models. To support the proposal of the present work, approximately 150 attributes were obtained, in the context of the financial institution itself, that inform about products marketed by the financial institution. The sample is made up of customer data from a large Brazilian financial institution that started a relationship between January and December 2019. We performed a horserace with some of the most popular machine learning algorithms that perform well on tabular data: Logistic Regression, decision trees, random forest and support vector machine. Two metrics were used to measure the predictive ability of the models: area under the ROC curve (AUC) and accuracy. The results obtained indicate that the model based on a support vector machine presented a result slightly superior to the other models, obtaining the best results in all the metrics used, such as accuracy, precision, recall, F-measure and Area Under the Curve (AUC). The attributes with the highest predictive capacity were also investigated, which, in general, revealed as good predictors the attributes that map the time of relationship, contracting of housing credit and declared monthly income. The results obtained in this work have the capability to assist in reversing the scenario of falling credit card contracts at the IF object of analysis, which goes against the market trend.