A procedure for detection of border communities using convolution techniques

Abstract

[EN] Many complex systems can be modeled by graphs and networks, [4]. In some problems, the study of communities allows quantitative and qualitative approaches and obtaining some knowledge about the structure of the graph and what it represents [1, 3, 5]. There is extensive literature in the study of communities, mostly focused on non-directed graphs [3, 5]. In our case we focus our work on the study of communities in directed graphs, weakly connected, with weights on the edges. In MME&HB 2016, was presented an algorithm for detection of directional communities in a directed graph [1], with a special interest in the graph representing the process of access to the Spanish Public University System, (SUPE) [1, 2]. The proposed algorithm allowed to obtain communities that provided an approximation to the problem. In MME&HB 2020 we propose a new algorithm based on obtaining the centers of the graph and pruning non-significant edges [6]. Recently a method for obtain communities using convolution techniques was presented in [5]. In this paper, we propose a community detection algorithm that combines a method for calculating potential community centers and convolution techniques to prune non-significant edges.