Modelos de classificação combinados a redes bipartidas em dados da NFL

Abstract

Complex networks have been widely used to depict real data, describing the interaction between objects through graphs, with great applicability in data analysis across various areas, such as airports, social relations, the internet World Wide Web, and sports. In this study, using data from the National Football League (NFL), networks are employed to represent the gameplay of a specific team, illustrating the dynamics of passes during a match and utilizing this information to extract outcomes and make predictions about games, such as in this study, the probability of one obtain a positive campaign in a season. One of the tools used for the analysis of complex networks is global measures, which resumes information about the entire network into a single value based on a particular characteristic and can in this way be incorporated into classification models as covariates. Therefore, global measures were used in classification models to attempt to explain the outcome of positive campaign in a NFL season.