Agrupación de instituciones financieras en Ecuador basada en el uso de tarjetas de crédito y débito usando técnicas de aprendizaje no supervisado

Abstract

CONDITION FOR PUBLICATION OF PROJECT. This thesis project analyzes the use of credit and debit cards in Ecuadorian financial institutions using unsupervised learning techniques, aiming to segment these entities based on their transactional behavior. Clustering algorithms such as K-Means and DBSCAN were employed, enabling the identification of usage patterns and the classification of banks into distinct groups. The data used comes from the Superintendency of Banks of Ecuador and includes detailed transactional information, which was preprocessed through normalization, feature selection, and dimensionality reduction. The models were then evaluated using metrics such as the silhouette coefficient and the elbow method, facilitating the identification of representative clusters. The results reveal four main segments: (1) institutions specialized in debit card transactions, (2) banks with low transactional activity, (3) leading credit institutions with high transaction volumes, and (4) diversified banks maintaining a balance between credit and debit operations. This segmentation provides valuable insights for strategic decision-making, financial product optimization, and risk management within the banking sector. The study concludes that clustering techniques offer a deeper understanding of the Ecuadorian financial ecosystem, supporting the development of commercial and regulatory strategies that promote inclusion and efficiency in the banking industry. Additionally, predictive analysis based on these groupings could contribute to the formulation of policies tailored to emerging trends in the financial market. Keywords: Clustering, Unsupervised learning, Finance, Segmentation, Credit and debit cards.