Estimação do número de comunidades em redes ponderadas

Abstract

The Stochastic Block Model is a commonly used model in real networks that exhibits community structure, that is, the vertices of the network are divided into groups. However, the number of communities related to the underlying model is not specified in real data, so it is necessary to use inferential methods to estimate this number. The objective is to study a method based on hypothesis tests that estimates the number of communities in binary networks, adapt it to weighted networks and perform a simulated analysis to comprehend the performance in this case. Furthermore, we applied hypothesis tests to estimate the number of communities and analyze the attribution of stations to communities in BART’s (Bay Area Rapid Transit) weighted network system, which is a rapid transit system that covers San Francisco’s bay.