Avaliação empírica de medidas de redes complexas para classificação de dados

Abstract

Complex networks are graphs with non-trivial connection structures, that is, they do not follow a regular or random pattern. Algorithms based on complex networks have gained increasing prominence in the classification of data, as they make it possible to interpret and model the pattern formation of the data and the relationships between them. As there are several measures of networks that have very different characteristics, the objective of this work is to characterize them, in the face of several problems, revealing more suitable scenarios to use one or another measure. In this sense, six measures of complex networks are investigated in this work from the classification via standard compliance. The results show that some of these measures have well-defined favorable scenarios for better predictive performance.