Agrupamento em conjunto de dados com múltiplos atributos via redução de dimensionalidade e detecção de comunidades

Abstract

This work falls within the field of Machine Learning and focuses on the challenge of Clustering in datasets with multiple variables. Addressing the issue of the Curse of Dimensionality, a common difficulty in high-dimensional data, the study focuses on Dimensionality Reduction and Community Detection techniques. The Curse of Dimensionality refers to the deterioration of the performance of certain learning algorithms as the dimensionality of the data increases. Few studies address the implications of graph-based clustering in this scenario, highlighting the need to investigate how clustering techniques can be adapted to effectively deal with high-dimensional datasets. Through the application of Community Detection Algorithms in conjunction with Graph Construction, the study aims to explore efficient methods for handling large volumes of data. The experiments conducted showed that, in datasets with a large number of samples, the use of Dimensionality Reduction techniques, in particular, resulted in superior performance in identifying significant clusters. This result emphasizes the importance of Dimensionality Reduction as a crucial tool for overcoming the challenges posed by the Curse of Dimensionality in Machine Learning environments. This work can assist future investigations and practical applications in complex and multidimensional data analysis.