A carregar...
Publicação
Identificação de perfis de clientes bancários : uma perspectiva de Ciência de Dados
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 5.09 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
The banking sector is essential for the functioning of society. With technological advancements, banks had to innovate their services to remain relevant. This trend arose due to the increasing competition, involving not only established banks but also new entities (neobanks, etc).
Customer data analysis is crucial for the growth and sustainability of any bank, as it allows for a
detailed understanding of customers, thereby enhancing their satisfaction and loyalty.
The field that develops this analysis is called Customer Intelligence (CI). For a given bank, CI
issues revolve around difficulties in determining the “value” of each customer or customer group,
identifying those posing higher financial risk, and personalizing the provided offers. Academically, studies conducted in CI are limited, both in their availability and the quantity of applied
methods, as well as in the clear presentation of conclusions for bank administrators responsible
for decision-making.
The objective of this work was to develop a robust and comprehensive methodology in an
exploratory manner, using demographic, transactional, and loan data from bank customers. The
culmination of this methodology was the creation of a dashboard that objectively highlights the results of essential conclusions to understand a given customer. This dashboard includes a summary
of the developed analyses, customer grouping, and characterization of customers.
The developed methodology focuses on four phases: 1) summarizing the transactional behavior of a customer based on RFM analysis, which resulted in six customer segments; 2) customer
clustering using the k-means algorithm, revealing the existence of three groups; 3) identification
of profiles for each group using data mining techniques; 4) customer classification using machine
learning algorithms such as XGBoost, which achieved the best performance.
Descrição
Tese de Mestrado, Ciência de Dados, 2024, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Customer intelligence Banca Algoritmos de agrupamento Análise de padrões Reportagem com dashboard Teses de mestrado - 2024