Aplicación del algoritmo k-prototypes para la segmentación de compañías ecuatorianas supervisadas por la superintendencia de compañías, valores y seguros

Abstract

This research focuses on the analysis and segmentation of Ecuadorian companies registered with the Superintendence of Companies, Securities, and Insurance (SuperCias), utilizing financial, sectoral, and geographical variables from the fiscal year 2024. Despite the significance of SuperCias in regulating and stabilizing the national corporate environment, it has been identified that the complexity and heterogeneity of the reported information hinder the detection of performance patterns through traditional statistical methods. This limitation restricts the ability to guide evidence-based decisions, resulting in an incomplete analysis of the sector. By employing unsupervised learning techniques, specifically the K-Prototypes algorithm within the Team Data Science Process (TDSP) framework, this study aims to identify natural groupings that capture both the categorical and numerical data present in the dataset. The implementation of this model allows for the simultaneous processing of quantitative financial metrics and qualitative corporate data. The segmentation results intend to reveal distinct business clusters, highlighting specific characteristics within each group. These strategies are designed to optimize the analysis of mixed-type datasets, providing a more robust understanding of the corporate landscape.