A Hybrid System for Financial Counselling in Fintech Lending Application


David Josué Barrientos Rojas
Marie Chantelle Cruz
João Fausto Lorenzato de Oliveira


In a world where high connectivity, portability, and speed have become usual demands in every service, traditional banking has been constantly challenged into evolving and finding better solutions through technology. At this point is where fintechs have gained a huge market portion, setting the pace for the future in banking.In this research project, the case of a specific fintech is considered: Justa, a Brazilian business intended to facilitate common banking services to traders, such as debit and credit transactions, loans, and accounts management. For this project a computational intelligence-based system is developed to attempt to predict accurately the possibility of a client of Justa succeeding at achieving a goal set at the very beginning of the partnership. This, based on their declared characteristics and stored information of past clients.The system was elaborated from a classification approach, considering the widely known benefits of hybrid systems. Various models were tested using parameter selection in preprocessed data, four of them were then picked to participate in a voting ensemble to make predictions: MLP, KNN, Decision Tree, and Naïve Bayes. Results for the hybrid classification system were by objective metrics: 0.609 accuracy, 0.845 recall, 0.577 precision, and 0.676; which indicate an improvement over the ones obtained in each individual model and align with Justa’s interests since reflect a system that is best at predicting true positives.Overall, the proposed system achieved satisfactory results with the given data and its limitations. However, it is considered a successful approach.


Não há dados estatísticos.


Como Citar
Barrientos Rojas, D., Cruz, M., & Oliveira, J. F. (2023). A Hybrid System for Financial Counselling in Fintech Lending Application. Revista De Engenharia E Pesquisa Aplicada, 7(3), 73-82. https://doi.org/10.25286/repa.v7i3.2462
Edição Especial em Ciência de Dados e Analytics