Role-based Label Propagation Algorithm for Community Detection

📝 Abstract

Community structure of networks provides comprehensive insight into their organizational structure and functional behavior. LPA is one of the most commonly adopted community detection algorithms with nearly linear time complexity. But it suffers from poor stability and occurrence of monster community due to the introduced randomize. We note that different community-oriented node roles impact the label propagation in different ways. In this paper, we propose a role-based label propagation algorithm (roLPA), in which the heuristics with regard to community-oriented node role were used. We have evaluated the proposed algorithm on both real and artificial networks. The result shows that roLPA is comparable to the state-of-the-art community detection algorithms.

💡 Analysis

Community structure of networks provides comprehensive insight into their organizational structure and functional behavior. LPA is one of the most commonly adopted community detection algorithms with nearly linear time complexity. But it suffers from poor stability and occurrence of monster community due to the introduced randomize. We note that different community-oriented node roles impact the label propagation in different ways. In this paper, we propose a role-based label propagation algorithm (roLPA), in which the heuristics with regard to community-oriented node role were used. We have evaluated the proposed algorithm on both real and artificial networks. The result shows that roLPA is comparable to the state-of-the-art community detection algorithms.

📄 Content

Corresponding author. Tel.: +86 13866151397. E-mail address: ahhfhw@163.com (W. He). Role-based Label Propagation Algorithm for Community Detection Xuegang Hu, Wei He, Huizong Li, Jianhan Pan School of Computer & Information, Hefei University of Technology, Hefei, 230009, China Abstract Community structure of networks provides comprehensive insight into their organizational structure and functional behavior. LPA is one of the most commonly adopted community detection algorithms with nearly linear time complexity. But it suffers from poor stability and occurrence of monster community due to the introduced randomize. We note that different community-oriented node roles impact the label propagation in different ways. In this paper, we propose a role-based label propagation algorithm (roLPA), in which the heuristics with regard to community-oriented node role were used. We have evaluated the proposed algorithm on both real and artificial networks. The result shows that roLPA is comparable to the state-of-the-art community detection algorithms. Keywords: Community-oriented; node role; heuristics; label propagation; community detection; complex networks. 1 Introduction Many real-world systems can be modeled as network, such as on-line social networks, scientist collaboration networks, epidemic networks, electric networks, the Internet, World Wide Web, and metabolic networks. Community structure [1] of network is the tendency for nodes to be assembled into groups, or communities, which densely connected inside and loosely connected with the rest of the network. For many real-world systems, community structure provides comprehensive insight into their organizational structure and functional behavior. For example, spreading of epidemic disease, which can be seen as a dynamic process on network, is greatly affected by the community structure of the social network [2]. Community detection algorithms, which aim to reveal hidden community structure in network, have attracted significant attention in recent few years. A substantial number of community detection algorithms have been proposed, including modularity optimization algorithms, spectral clustering algorithms, hierarchical partition algorithms, and information theory based algorithms [3]. Among them, the label propagation algorithm (LPA) [4] proposed by Raghavan et al. is one of the most commonly adopted algorithms with nearly linear time complexity. Owing to its prominent speed, conceptual simplicity, accurateness, and easy implementation of parallelism, LPA is suitable for large-scale networks with millions of nodes and edges, e.g. Facebook, which had almost 1.4 billion monthly active users by the end of 2014. Despite various advantages, some issues of the LPA have not been properly addressed, such as the deficiency of robustness and stability due to its random updating and breaking tie strategy. The weak robustness of the LPA means the solutions in different runs can be quite different. And that is related to the significance of community structure in a network, that is, for a network with weaker community structure, the LPA would get the less stable solution. Also for a network with weak community structure, the LPA may produce a monster community, or giant community, which dominate a big part of the network and swallow the small communities, making the solution trivial [5]. Raghavan et al. point out that, if the core nodes of network are identified, the LPA can be implemented by initializing core nodes with unique labels and leaving the other nodes unlabeled [4]. In this case the unlabeled nodes will have a tendency to acquire labels from their closest attractor and join that community. However, this implementation attracted little attention due to its inherent limitation, that is, identifying the core node of community before identifying the community itself is very difficult. On the other hand, identifying the core nodes without considering community structure may lead the algorithm toward an undesired solution. For example, the core nodes identified through the degree centrality may be the inter-community hubs, e.g. overlapping nodes with high degree shared by two or more communities, which would lead the algorithm to failure. In this paper, we propose a role-based label propagation algorithm (roLPA) by introducing the heuristics inspired by community-oriented node role, and defined some new measure for the node preference by considering the community-oriented role. We apply the algorithm on both real and artificial networks. The algorithm is shown to be comparable to the state-of-the- art community detection algorithms, with nearly linear time complexity. The rest part of the paper is arranged as follows. In Section 2 we review label propagation algorithm and community-oriented node role. In Section 3 we stated the proposed algorithm. In Section 4, we present the result of experiments. Section

This content is AI-processed based on ArXiv data.