Community detection algorithms: a comparative analysis: invited presentation, extended abstract

Detecting the community structure exhibited by real networks is a crucial step toward an understanding of complex systems beyond the local organization of their constituents. Many algorithms proposed so far, especially the group of methods in the k-means formulation, can lead to a high degree of efficiency and accuracy. Here we test three k-means methods, based on optimal prediction, diffusion distance and dissimilarity index, respectively, on two artificial networks, including the widely known ad hoc network with same community size and a recently introduced LFR benchmark graphs with heterogeneous distributions of degree and community size. All of them display an excellent performance, with the additional advantage of low computational complexity, which enables the analysis of large systems. Moreover, successful applications to several real world networks confirm the capability of the methods.

As an open issue, the measure of community influential is no uniform standard, how to measure the influence of community has attracted extensive attention. This paper proposed a quantitative measure to identify the influence of community. Based on the state of critical functionality (SCF), a new function, which names as the weighted state of critical functionality (WSCF), is defined. For the WSCF, not only the connections among communities but also the topology within community is considered. When the community structure of the complex network is divided, each community is renormalized as a node by the renormalization method. Then, the influence of community is measured by the values of WSCF, the greater the value of WSCF, the less the influence of the corresponding community. The influence of community of three classic constructed networks (i.e., a Erodös-Rényi (ER) random network, a BA scale free network and a small-word (SW) network) is measured by the proposed method. To further verify the feasibility of the method, two community detection algorithms are used to divide community structure in the real networks. The influence of the community of the 9/11 terrorist network, a US Air network and a PolBooks network is measured by the proposed method. The influence of each community could be measured and the most influential community in each network is identified by the proposed method. The results reveal that the proposed method is a feasible measure to identify the influence of community, its recognition effect is better than SCF, and accuracy is higher. The SCF is a special case of WSCF, when the weights and cluster coefficients are equal to 1. For the proposed method, once the community structure of the network is divided, the corresponding community influence is identified by the proposed method, which is not affected by the community division algorithms.

With the development of network technology and the continuous advancement of society, the combination of various industries and the Internet has produced many large-scale complex networks. A common feature of complex networks is the community structure, which divides the network into clusters with tight internal connections and loose external connections. The community structure reveals the important structure and topological characteristics of the network. The detection of the community structure plays an important role in social network analysis and information recommendation. Therefore, based on the relevant theory of complex networks, this paper introduces several common community detection algorithms, analyzes the principles of particle swarm optimization (PSO) and genetic algorithm and proposes a particle swarm-genetic algorithm based on the hybrid algorithm strategy. According to the test function, the single and the proposed algorithm are tested, respectively. The results show that the algorithm can maintain the good local search performance of the particle swarm optimization algorithm and also utilizes the good global search ability of the genetic algorithm (GA) and has good algorithm performance. Experiments on each community detection algorithm on real network and artificially generated network data sets show that the particle swarm-genetic algorithm has better efficiency in large-scale complex real networks or artificially generated networks.

We examine a community structure in random graphs of size n and link probability p/n determined with the Newman greedy optimization of modularity. Calculations show that for p<1 communities are nearly identical with clusters. For p=1 the average sizes of a community s(av) and of the giant community s(g) show a power-law increase s(av)∼n(α') and s(g)∼n(α). From numerical results we estimate α'≈0.26(1) and α≈0.50(1) and using the probability distribution of sizes of communities we suggest that α'=α/2 should hold. For p>1 the community structure remains critical: (i) s(av) and s(g) have a power-law increase with α'≈α<1 and (ii) the probability distribution of sizes of communities is very broad and nearly flat for all sizes up to s(g). For large p the modularity Q decays as Q∼p(-0.55), which is intermediate between some previous estimations. To check the validity of the results, we also determine the community structure using another method, namely, a nongreedy optimization of modularity. Tests with some benchmark networks show that the method outperforms the greedy version. For random graphs, however, the characteristics of the community structure determined using both greedy and nongreedy optimizations are, within small statistical fluctuations, the same.

Network models with latent geometry have been used successfully in many applications in network science and other disciplines, yet it is usually impossible to tell if a given real network is geometric, meaning if it is a typical element in an ensemble of random geometric graphs. Here we identify structural properties of networks that guarantee that random graphs having these properties are geometric. Specifically we show that random graphs in which expected degree and clustering of every node are fixed to some constants are equivalent to random geometric graphs on the real line, if clustering is sufficiently strong. Large numbers of triangles, homogeneously distributed across all nodes as in real networks, are thus a consequence of network geometricity. The methods we use to prove this are quite general and applicable to other network ensembles, geometric or not, and to certain problems in quantum gravity.

Community detection has become a very important part in complex networks analysis. Authors traditionally test their algorithms on a few real or artificial networks. Testing on real networks is necessary, but also limited: the considered real networks are usually small, the actual underlying communities are generally not defined objectively, and it is not possible to control their properties. Generating artificial networks makes it possible to overcome these limitations. Until recently though, most works used variations of the classic Erdős-Rényi random model and consequently suffered from the same flaws, generating networks not realistic enough. In this work, we use Lancichinetti et al. model, which is able to generate networks with controlled power-law degree and community distributions, to test some community detection algorithms. We analyze the properties of the generated networks and use the normalized mutual information measure to assess the quality of the results and compare the considered algorithms.

Complex networks are composed of nodes and complex relationships between nodes, which are the abstraction of various complex systems in the real world. Community structure is a remarkable characteristic of complex networks.Community detection algorithms can reveal the topological structure of the networks and mine the potential information in complex networks, which can be widely used in personalized recommendation, public security, social science and other fields, and has important research significance. The community detection algorithm based on seeds extension is a classic community detection algorithm, which has high execution efficiency, but there are some problems. Aiming at the problems of poor seed nodes selection and improper community expansion method of seeds expansion algorithm, this paper proposes a seed expansion algorithm based on CRITIC method and node gravity which is called SECNG. The contributions of the algorithm are as follows: At first, we introduced CRITIC method to measure the importance of nodes, which can objectively measure the importance of nodes in the networks. Secondly, we propose node gravity and community gravity to guide the expansion of communities. In addition, modularity maximization strategy is used to merge communities. We verify the effectiveness of the SECNG algorithm through the real networks and artificial generated networks. The results show that the Q value and NMI values of the algorithm are higher than other algorithms.

Modularity is widely used to effectively measure the strength of the disjoint community structure found by community detection algorithms. Although several overlapping extensions of modularity were proposed to measure the quality of overlapping community structure, there is lack of systematic comparison of different extensions. To fill this gap, we overview overlapping extensions of modularity to select the best. In addition, we extend the Modularity Density metric to enable its usage for overlapping communities. The experimental results on four real networks using overlapping extensions of modularity, overlapping modularity density, and six other community quality metrics show that the best results are obtained when the product of the belonging coefficients of two nodes is used as the belonging function. Moreover, our experiments indicate that overlapping modularity density is a better measure of the quality of overlapping community structure than other metrics considered.

The complexity of many community detection algorithms is usually an exponential function with the scale which hard to uncover community structure with high speed. Inspired by the ideas of the famous modularity optimization, in this paper, we proposed a proper weighting scheme utilizing a novel k-strength relationship which naturally represents the coupling distance between two nodes. Community structure detection using a generalized weighted modularity measure is refined based on the weighted k-strength matrix. We apply our algorithm on both the famous benchmark network and the real networks. Theoretical analysis and experiments show that the weighted algorithm can uncover communities fast and accurately and can be easily extended to large-scale real networks.

To unravel the driving patterns of networks, the most popular models rely on community detection algorithms. However, these approaches are generally unable to reproduce the structural features of the network. Therefore, attempts are always made to develop models that incorporate these network properties beside the community structure. In this article, we present a probabilistic generative model and an efficient algorithm to both perform community detection and capture reciprocity in networks. Our approach jointly models pairs of edges with exact two-edge joint distributions. In addition, it provides closed-form analytical expressions for both marginal and conditional distributions. We validate our model on synthetic data in recovering communities, edge prediction tasks and generating synthetic networks that replicate the reciprocity values observed in real networks. We also highlight these findings on two real datasets that are relevant for social scientists and behavioural ecologists. Our method overcomes the limitations of both standard algorithms and recent models that incorporate reciprocity through a pseudo-likelihood approximation. The inference of the model parameters is implemented by the efficient and scalable expectation–maximization algorithm, as it exploits the sparsity of the dataset. We provide an open-source implementation of the code online.

Community structure analysis is a powerful tool for social networks that can simplify their topological and functional analysis considerably. However, since community detection methods have random factors and real social networks obtained from complex systems always contain error edges, evaluating the significance of a partitioned community structure is an urgent and important question. In this paper, integrating the specific characteristics of real society, we present a framework to analyze the significance of a social community. The dynamics of social interactions are modeled by identifying social leaders and corresponding hierarchical structures. Instead of a direct comparison with the average outcome of a random model, we compute the similarity of a given node with the leader by the number of common neighbors. To determine the membership vector, an efficient community detection algorithm is proposed based on the position of the nodes and their corresponding leaders. Then, using a log-likelihood score, the tightness of the community can be derived. Based on the distribution of community tightness, we establish a connection between p-value theory and network analysis, and then we obtain a significance measure of statistical form . Finally, the framework is applied to both benchmark networks and real social networks. Experimental results show that our work can be used in many fields, such as determining the optimal number of communities, analyzing the social significance of a given community, comparing the performance among various algorithms, etc.

The amount of data generated by online social networks such as Facebook, Twitter, etc., has recently experienced an enormous growth. Extracting useful information such as community structure, from such large networks is very important in many applications. Community is a collection of nodes, having dense internal connections and sparse external connections. Community detection algorithms aim to group nodes into different communities by extracting similarities and social relations between nodes. Although, many community detection algorithms in literature, they are not scalable enough to handle large volumes of data generated by many of the today's big data applications. So, researchers are focusing on developing parallel community detection algorithms, which can handle networks consisting of millions of edges and vertices. In this article, we present a comprehensive survey of parallel community detection algorithms, which is the first ever survey in this domain, although, multiple papers exist in literature related to sequential community detection algorithms.

Proper testing of graph mining algorithms, for example, algorithms for community detection, requires the capability of creating realistic random graphs. As our understanding of real graph communities evolves, so should the random graph generators evolve, too. In this work, we propose a random graph generator called HyGen that, unlike the existing random graph generators, is designed to preserve the community structure of real networks, especially the commonly observed hyperbolic intra-community connectivity structure. The generated graphs will also preserve the total degree distributions and clustering coefficients of the original graph without introducing too much determinism. In addition, we also propose realistic distributions for the parameters controlling the hyperbolic shape of the communities.

Overlapping community is a phenomenon often observed in numerous real-world networks. Fire-spread (Pattanayak et al. in Swarm Evol Comput. 44: 1–48 (2019) [11]) community detection algorithm is an efficient algorithm to detect overlapping community structures. In this work, the Fire-spread algorithm is modified to establish a relationship between community structure and clustering coefficient. By using different networks and executing the modified Fire-spread algorithm, it is found that the clustering coefficient is highly correlated with community structure. Finally, a simpler community detection algorithm, derived from the fire-spread algorithm, is proposed, where the clustering coefficient is used as a threshold value. To validate the proposed algorithm, it is compared with some state of art community detection algorithms based on the NMI score.

Community detection is one of the primary tools to discover useful information that is hidden in complex networks. Some community detection algorithms for bipartite networks have been proposed from various viewpoints. However, the performance of these algorithms deteriorates when the community structure becomes unclear. Enhancing community structure remains a nontrivial task. In this paper, we propose a community detection algorithm, called ECD, that enhances community structure in bipartite networks. In the proposed ECD, the topology of a network is modified by reducing unnecessary edges that are connected to neighboring low-weight communities. Therefore, an ambiguous community structure is converted into a structure that is much clearer than the original structure. The experimental results on both artificial and real-world networks verify the accuracy and reliability of our algorithm. Compared with existing community detection algorithms using state-of-the-art methods, our algorithm has better performance.