Autonomous overlapping community detection in temporal networks: A dynamic Bayesian nonnegative matrix factorization approach

Abstract

A wide variety of natural or artificial systems can be modeled as time-varying or temporal networks.To understand the structural and functional properties of these time-varying networked systems, it is desirable to detect and analyze the evolving community structure. In temporal networks, the identified communities should reflect the current snapshot network, and at the same time be similar to the communities identified in history or say the previous snapshot networks. Most of the existing approaches assume that the number of communities is known or can be obtained by some heuristic methods. This is unsuitable and complicated for most real world networks, especially temporal networks. In this paper, we propose a Bayesian probabilistic model, named Dynamic Bayesian Nonnegative Matrix Factorization (DBNMF), for automatic detection of overlapping communities in temporal networks. Our model can not only give the overlapping community structure based on the probabilistic memberships of nodes in each snapshot network but also automatically determines the number of communities in each snapshot network based on automatic relevance determination. Thereafter, a gradient descent algorithm is proposed to optimize the objective function of our DBNMF model. The experimental results using both synthetic datasets and real-world temporal networks demonstrate that the DBNMF model has superior performance compared with two widely used methods, especially when the number of communities is unknown and when the network is highly sparse.

A Survey of Community Detection Approaches: From Statistical Modeling to Deep Learning

“…Dynamic DBNMF (2016) [83] Present a Bayesian probabilistic model based on NMF to identify overlapping communities in temporal networks.…”

Change point detection in social networks—Critical review with experiments

“…In the first case, groups are identified specifically from members of the graph exhibiting common connective behaviour while a change is identified when groups or nodes behaviours no longer conform to the group structures. Of course, a number of approaches have been proposed to detect dynamics of the communities and their evolution over time [19], [20], [21], and [22]. Many of the community detection methods find groups by optimizing selected metric (e.g.…”

“…Now research is increasingly viewing networks as dynamic systems, where the dynamic properties are as important as overall network structure. The computational capability to study not only large graphs, but a long sequence of large graphs over time has led to growing research in the field of detecting, modelling and predicting changes in complex networks [1,2,3,4,5,6,7]. The focus of this paper is on the problem of change point detection, which is a form of dynamic anomaly detection that has a long history of study in traditional time series datasets [8,9,10,11,12,13].…”

Community Detection of Heterogeneous Networks with RankClus Based on Spark GraphX

“…Because of the characteristics of multi-dimensional and multimode in the heterogenous, some researchers use data reconstruction method [9][10] and dimensionality reduction method [11][12] [13] to convert it into relatively simple network type or reduce the dimension. Liu et al [9] transform an original heterogeneous network into a bipartite network and perform community detection on it.…”

“…Each node and edge or hyperedge in the original heterogeneous network is, respectively, mapped into a vertex node and a link node in the bipartite network. Wang et al [11] propose a Bayesian probabilistic model, DBNMF model, for automatic detection of overlapping communities in temporal networks. Yang et al [13] present BIGCLAM model, an overlapping community detection method that scales to large networks of millions of nodes and edges.…”