Análisis de técnicas de clusterización aplicados a la segmentación de datos financieros

Abstract

Abstract: In the last decade, there has been a growing demand from financial and commercial institutions to adopt a customer-centric approach. This has generated the need to segment customers to provide specialized services tailored to their individual profiles. While numerous research studies have analyzed segmentation algorithms in fields like medicine and the environment, there is a scarcity of work focused on financial data, which is notoriously diverse and challenging to analyze due to its heterogeneous nature. In this research, three datasets from Kaggle were selected, and four segmentation algorithms were evaluated: K-means, Mean-Shift, Hierarchical Agglomerative, and DBSCAN. These algorithms underwent three experiments to assess their performance. The results indicate that under normal conditions, the K-means and Hierarchical Agglomerative algorithms excel in performance, while Mean-Shift and DBSCAN can outperform them under controlled circumstances. However, all algorithms show lower performance when dealing with highly dimensional datasets, with DBSCAN being the most affected by this characteristic.