DyPerm: Maximizing Permanence for Dynamic Community Detection

Abstract

In this paper, we propose DyPerm, the first dynamic community detection method which optimizes a novel community scoring metric, called permanence. DyPerm incrementally modifies the community structure by updating those communities where the editing of nodes and edges has been performed, keeping the rest of the network unchanged. We present strong theoretical guarantees to show how/why mere updates on the existing community structure leads to permanence maximization in dynamic networks, which in turn decreases the computational complexity drastically. Experiments on both synthetic and six real-world networks with given ground-truth community structure show that DyPerm achieves (on average) 35% gain in accuracy (based on NMI) compared to the best method among four baseline methods. DyPerm also turns out to be 15 times faster than its static counterpart. 1 final communities [16,23]. In this paper, we propose DyPerm, the first dynamic community detection method that adopts an effective community goodness metric, called "permanence" [12,13,9] and optimizes it to incrementally detect the community structure. The benefits of adopting permanence as an optimization function are two-fold: (i) Permanence, being a local vertex-centric metric (as opposed to the global network-centric metrics such as modularity, conductance), allows us to reassign communities to only those nodes whose associated topological structure has changed, and guarantees that the remaining nodes do not affect the optimization. This leads to very low computing complexity in updating the community structure when the network changes dynamically. (ii) Incremental changes in the local portion of the community structure guarantee that the resultant communities are highly correlated with that in the previous time-stamp. We present theoretical justifications why/how mere changes in the community structure lead to maximize permanence.We experiment with both synthetic and six real-world dynamic networks with known ground-truth community structure. A thorough comparative evaluation with four state-of-the-art baseline methods shows that DyPerm significantly outperforms all the baselines across different networks -DyPerm achieves up to 35% improvement in terms of Normalized Mutual Information (NMI) w.r.t. the best baseline method. Moreover, DyPerm tunrs out to be extremely fast, achieving up to 15 times speedup w.r.t. its static counterpart. In short, DyPerm is a fast and accurate dynamic community detection method.

Identifying Communities in Dynamic Networks Using Information Dynamics

“…DyPerm [29] is an optimization method based on permanence, which is also an incremental dynamic community detection method. DyPerm needs to specify the actual communities at the beginning.…”

“…One advantage of the QCA approach is the low time complexity due to the incremental technology it employs. Agarwal et al [29] introduced a community detection method (DyPerm) based on maximizing permanence in dynamic networks. One disadvantage of the DyPerm method is that it needs to have the initial community structure of the network, but this is unknown in many real networks.…”

Community Detection in Dynamic Social Networks: A Multi-Agent Aya based on Electric Field

“…Recently, some approaches for the detection of dynamic communities have been proposed [16], [17], [18], [19], [20], [5], [21], [22], [23], [24]. One way to analyze communities in an evolving network is to consider the dynamic graph as a succession of independent captures of the graph, all of which are static graphs.…”

“…So the evolution of the network is no longer considered as a succession of snapshots, but as a succession of modifications on the network. The idea is to start with an initial partition and to update it according to the latest evolution of the network instead of finding a new partition [28], [29], [23], [24]. The detection of communities is therefore not done on the whole network, but only by minor and successive local modifications.…”

Temporal Label Walk for Community Detection and Tracking in Temporal Network

“…Li et al divide communities according to permanence index [6]. Agarwal et al maximize permanence to detect communities [7]. Guo et al focus on local interact for community detection [8].…”