Aprendizaje no supervisado en el perfilamiento de clientes para profit scoring: caso de estudio de una Fintech Latinoamericana

The microcredit business has become popular, especially with the rise of Fintech companies. However, risk assessment of this type of credit remains a challenge given the uniqueness of its target clients. From this perspective, for the fintech lenders, it is not only important to know the probability of default but also the potential profitability of the loan, in order to make the decision as to whether or not to grant the microcredit. This project focuses on the development of a customer characterization and profiling model to make a profit score for the case study of a Latin American fintech firm. For this, unsupervised machine learning techniques are used, specifically a hybrid clustering method based on Kohonen self-organizing maps and K-means algorithms is presented. From the resulting groups, clients are labeled as undesirable or desirable. For the latter, a profile is made that define their most relevant characteristics. In this way, the model developed proves to be a key tool both for the decision making of approval or denial of credit and for the characterization of desirable clients, which supports decision-making in other fundamental areas of the business such as marketing and placement.

• Autores: Moya Gutiérrez, María Juliana
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2020
• Institución: Universidad de los Andes
• Repositorio: Séneca: repositorio Uniandes
• Idioma: spa
• OAI Identifier: oai:repositorio.uniandes.edu.co:1992/51481
• Acceso en línea: http://hdl.handle.net/1992/51481
• Palabra clave: Microfinanzas; Comportamiento del consumidor; Aprendizaje automático (Inteligencia artificial); Ingeniería
• Rights: openAccess
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/