Comparação de Técnicas de Agrupamento: K-means, Método Hierárquico de Ward e Misturas de Gaussianas Aplicadas a Dados Reais

Abstract

In the context of unsupervised learning, clustering methods are essential for data analysis, allowing the organization of elements into groups (or clusters) based on similar characteristics. This study aims to explore and compare three widely used clustering methods in the statistical literature: K-means, Ward’s method, and Gaussian mixtures. Although all methods share the goal of cluster formation, they differ significantly in their approaches and underlying assumptions. Using real datasets, we applied each of the aforementioned methods, evaluating their performance and effectiveness in group formation. We concluded that the choice of clustering method should consider the intrinsic properties of the data and the analysis objectives. This study contributes to the understanding of the differences and similarities between these methods, providing practical guidelines for applications in various data analysis scenarios.