Improved Community Detection Research Articles

Embedding Both Finite and Infinite Communities on Graphs [Application Notes

In this paper, we introduce a new setting for graph embedding, which considers embedding communities instead of individual nodes. We find that community embedding is not only useful for community-level applications such as graph visualization but also provide an exciting opportunity to improve community detection and node classification. Specifically, we consider the interaction between community embedding and detection as a closed loop, through node embedding. On the one hand, node embedding can improve community detection since the detected communities are used to fit a community embedding. On the other hand, community embedding can be used to optimize node embedding by introducing a community-aware high-order proximity. However, in practice, the number of communities can be unknown beforehand; thus we extend our previous Community Embedding (ComE) model. We propose ComE+, a new model which handles both: the unknown truth community assignments and the unknown number of communities present in the dataset. We further develop an efficient inference algorithm for ComE+ for keeping complexity low. Our extensive evaluation shows that ComE+ improves the state-of-the-art baselines in various application tasks, e.g., community detection and node classification.

Incidence of eclampsia and related complications across 10 low- and middle-resource geographical regions: Secondary analysis of a cluster randomised controlled trial.

BackgroundIn 2015, approximately 42,000 women died as a result of hypertensive disorders of pregnancy worldwide; over 99% of these deaths occurred in low- and middle-income countries. The aim of this paper is to describe the incidence and characteristics of eclampsia and related complications from hypertensive disorders of pregnancy across 10 low- and middle-income geographical regions in 8 countries, in relation to magnesium sulfate availability.Methods and findingsThis is a secondary analysis of a stepped-wedge cluster randomised controlled trial undertaken in sub-Saharan Africa, India, and Haiti. This trial implemented a novel vital sign device and training package in routine maternity care with the aim of reducing a composite outcome of maternal mortality and morbidity. Institutional-level consent was obtained, and all women presenting for maternity care were eligible for inclusion. Data on eclampsia, stroke, admission to intensive care with a hypertensive disorder of pregnancy, and maternal death from a hypertensive disorder of pregnancy were prospectively collected from routine data sources and active case finding, together with data on perinatal outcomes in women with these outcomes. In 536,233 deliveries between 1 April 2016 and 30 November 2017, there were 2,692 women with eclampsia (0.5%). In total 6.9% (n = 186; 3.47/10,000 deliveries) of women with eclampsia died, and a further 51 died from other complications of hypertensive disorders of pregnancy (0.95/10,000). After planned adjustments, the implementation of the CRADLE intervention was not associated with any significant change in the rates of eclampsia, stroke, or maternal death or intensive care admission with a hypertensive disorder of pregnancy. Nearly 1 in 5 (17.9%) women with eclampsia, stroke, or a hypertensive disorder of pregnancy causing intensive care admission or maternal death experienced a stillbirth or neonatal death. A third of eclampsia cases (33.2%; n = 894) occurred in women under 20 years of age, 60.0% in women aged 20–34 years (n = 1,616), and 6.8% (n = 182) in women aged 35 years or over. Rates of eclampsia varied approximately 7-fold between sites (range 19.6/10,000 in Zambia Centre 1 to 142.0/10,000 in Sierra Leone). Over half (55.1%) of first eclamptic fits occurred in a health-care facility, with the remainder in the community. Place of first fit varied substantially between sites (from 5.9% in the central referral facility in Sierra Leone to 85% in Uganda Centre 2). On average, magnesium sulfate was available in 74.7% of facilities (range 25% in Haiti to 100% in Sierra Leone and Zimbabwe). There was no detectable association between magnesium sulfate availability and the rate of eclampsia across sites (p = 0.12). This analysis may have been influenced by the selection of predominantly urban and peri-urban settings, and by collection of only monthly data on availability of magnesium sulfate, and is limited by the lack of demographic data in the population of women delivering in the trial areas.ConclusionsThe large variation in eclampsia and maternal and neonatal fatality from hypertensive disorders of pregnancy between countries emphasises that inequality and inequity persist in healthcare for women with hypertensive disorders of pregnancy. Alongside the growing interest in improving community detection and health education for these disorders, efforts to improve quality of care within healthcare facilities are key. Strategies to prevent eclampsia should be informed by local data.Trial registrationISRCTN: 41244132.

A Link-Based Similarity for Improving Community Detection Based on Label Propagation Algorithm

Community structure is one of the most best-known properties of complex networks. Finding communities help us analyze networks from a mesoscopic viewpoints instead of microscopic or macroscopic one. It helps to understand behavior grouping. Various community detection algorithms have been proposed with some shortcomings in time and space complexity, accuracy, or stability. Label Propagation Algorithm (LPA) is a popular method used for finding communities in an almost-linear time-consuming process. However, its performance is not satisfactory in some metrics such as accuracy and stability. In this paper, a new modified version of LPA is proposed to improve the stability and accuracy of the LPA by defining two concepts -nodes and link strength based on semi-local similarity-, while preserving its simplicity. In the proposed method a new initial node selection strategy, namely the tiebreak strategy, updating order and rule update are presented to solve the random behavior problem of original LPA. The proposed algorithm is evaluated on artificial and real networks. The experiments show that the proposed algorithm is close to linear time complexity with better accuracy than the original LPA and other compared methods. Furthermore, the proposed algorithm has the robustness and stability advantages while the original LPA does not have these features.

Leveraging the nonuniform PSO network model as a benchmark for performance evaluation in community detection and link prediction

Advances in network geometry pointed out that structural properties observed in networks derived from real complex systems can emerge in the hyperbolic space (HS). The nonuniform popularity-similarity-optimization (nPSO) is a generative model recently introduced in order to grow random geometric graphs in the HS, reproducing networks that have realistic features such as high clustering, small-worldness, scale-freeness and rich-clubness, with the additional possibility to control the community organization. Generative models allowing to tune the structural properties of ‘realistic’ synthetic networks are fundamental, because they offer a ground truth to investigate how predictive algorithms react to controlled topological variations. Here, we discuss how to leverage the nPSO model as a synthetic benchmark to compare the performance of methods for community detection and link prediction; and we prove that the nPSO offers a reliable and realistic testing framework which can complement other existing benchmarks not based on latent geometry. Furthermore, we confirm that network embedding information can improve community detection, whereas boosting link prediction in HS still needs further investigations. Indeed, we find that the presence of communities in nPSO significantly modifies the performance of link predictors and is fundamental for the reproducibility of results observed on real networks. The nPSO can trigger valuable insights to understand the intrinsic rules of link-growth and self-organization that connect topology to geometry and that are encoded in link prediction algorithms differentiating their performance.

Using triangles and latent factor cosine similarity prior to improve community detection in multi‐relational social networks

SummaryCommunity detection is a key to understanding the structure of complex networks. Communities, or clusters, are groups of vertices having higher probability of being connected to each other than to the members in other groups. Considering the importance of triangle structures, we first propose σ‐tensor to model ordinary relationships and triangle relationships simultaneously. Then, we propose a simple but effective latent factor prior, ie, latent factor cosine similarity prior, to improve community detection. The latent factor cosine similarity prior is a kind of statistics of the well‐defined synthetic multi‐relational social networks. It is based on a key observation that most latent feature factors of intra‐group members in these networks are highly similar according to cosine similarity measure. Using this prior along with the RESCAL tensor factorization model, we can obtain a superior latent feature factor matrix. Moreover, N‐RESCAL model, a variant of RESCAL model, and its corresponding algorithm N‐RESCAL‐ALS are proposed for the simplicity and the removal of the limit of cosine similarity. Once the latent factor matrix is obtained by factorizing σ‐tensor using N‐RESCAL model, we apply agglomerative clustering algorithm for community discovery. We call this framework as TNRA. Experiment results on several real‐world datasets are surprisingly promising, clearly demonstrating the power of the proposed prior and the effectiveness of our proposed methods.

Interest-Based Clustering Approach for Social Networks

Recently, the applications of community detection have increased because of their effectiveness in identifying communities correctly. Many methods and algorithms have been introduced to bring new insights that will improve community detection in social networks. While such algorithms can find useful communities, they tend to focus on network structure and ignore node interests and interconnections. However, accurate community detection requires the consideration of both network structure and node interests. The best method to achieve this is by utilizing unsupervised models. In this article, we introduce a new approach for social network clustering, termed Interest-based Clustering, which clusters nodes in social networks based on a measure of interest similarity. It considers structure, interaction, and node interest along with nodes friends’ interests. The empirical evaluation of this new approach was done using real dataset crawled from Twitter. The approach outperforms well-known community detections algorithms, SCAN, Fast Modularity, Zhao et al., in terms of modularity, connectivity, and overlapping.

Common strategy to improve community detection performance based on the nodes’ property

Improving the community detection algorithm is of great importance. The authors propose a novel method based on the nodes’ property in order to detect the community structure. Given a detected community structure, in which nodes have their community signals, the value of the modularity can be changed if a node's community sign change to other communities’ signs. Accordingly, the new method readjusts the affiliation between a node and its community in order to raise the modularity value. Experimental results of the detection for a list of open-source networks show that the proposed algorithm can detect better community structure than classic methodologies based on modularity.

Erratum to: Characterizing the structure of large real networks to improve community detection

Erratum to: Characterizing the structure of large real networks to improve community detection

Characterizing the structure of large real networks to improve community detection

Community structure is a fundamental feature for many networks. The problem of discovering communities in those networks thus has been attracting a lot of research. However, due to the rapid increase of networks’ scale and the availability of real communities in many networks, the task of detecting communities in large real networks remains a challenging problem. In this paper, we study the structure of various large real networks and their ground-truth community structures and observe an interesting phenomenon: the difference of degrees (abbreviated as dod) of connected nodes follows a heavy-tailed distribution with an approximate power-law tail for large dod, in both original network and community structure but to different extents. With the aim to explore the effect of this observation on identifying communities in real large networks, we propose a weighting strategy and further embed it into two prominent community detection algorithms. Comparisons against the state of the arts demonstrate a very promising performance of the proposed weighting strategy.

Improved community detection in weighted bipartite networks.

Real-world complex networks are composed of non-random quantitative interactions. Identifying communities of nodes that tend to interact more with each other than the network as a whole is a key research focus across multiple disciplines, yet many community detection algorithms only use information about the presence or absence of interactions between nodes. Weighted modularity is a potential method for evaluating the quality of community partitions in quantitative networks. In this framework, the optimal community partition of a network can be found by searching for the partition that maximizes modularity. Attempting to find the partition that maximizes modularity is a computationally hard problem requiring the use of algorithms. QuanBiMo is an algorithm that has been proposed to maximize weighted modularity in bipartite networks. This paper introduces two new algorithms, LPAwb+ and DIRTLPAwb+, for maximizing weighted modularity in bipartite networks. LPAwb+ and DIRTLPAwb+ robustly identify partitions with high modularity scores. DIRTLPAwb+ consistently matched or outperformed QuanBiMo, while the speed of LPAwb+ makes it an attractive choice for detecting the modularity of larger networks. Searching for modules using weighted data (rather than binary data) provides a different and potentially insightful method for evaluating network partitions.

Targeted revision: A learning-based approach for incremental community detection in dynamic networks

Community detection is a fundamental task in network analysis. Applications on massive dynamic networks require more efficient solutions and lead to incremental community detection, which revises the community assignments of new or changed vertices during network updates. In this paper, we propose to use machine learning classifiers to predict the vertices that need to be inspected for community assignment revision. This learning-based targeted revision (LBTR) approach aims to improve community detection efficiency by filtering out the unchanged vertices from unnecessary processing. In this paper, we design features that can be used for efficient target classification and analyze the time complexity of our framework. We conduct experiments on two real-world datasets, which show our LBTR approach significantly reduces the computational time while keeping a high community detection quality. Furthermore, as compared with the benchmarks, we find our approach’s performance is stable on both growing networks and networks with vertex/edge removals. Experiments suggest that one should increase the target classification precision while keeping recall at a reasonable level when implementing our proposed approach. The study provides a unique perspective in incremental community detection.

A new Pre-processing Strategy for Improving Community Detection Algorithms

In the study of complex networks, a network is said to have community structure if it divides naturally into groups of nodes with dense connections within groups and only sparser connections between them. Detecting communities from complex networks has attracted attention of researchers in a wide range of research areas, from biology to sociology and computer science. In this paper, we introduce a new approach to make existing community detection algorithms execute with better results. Our method enhancing community detection algorithms by applying a pre-processing step that exploits betweenness for nodes and edges, to maps unweighted graph onto a weighted graph. It has been tested in conjunction with four algorithms, namely the Louvain method, SOM algorithm, VOS clustering, and Danon algorithm. Experimental results show that our edge weighting strategies raises modularity for existing algorithms.

An Improved Random Walk Based Community Detection Algorithm

Community detection is an important issue in social network analysis, which aims at finding potential community structures such that the internal nodes of a community have higher closeness than external nodes. Taking into account node attribute information, this paper presents an improved community detection algorithm based on random walk. Based on the basic understanding that people getting together often relies on their common interests, node similarities are initially calculated with node attributes and iteratively updated based on the random walk model. Meanwhile, node importance is computed to represent how much it can influence other nodes, based on which some important nodes are selected as seeds for community clustering. As for overlapping community detection, some construction is made on a given social network. Experimental results on several real datasets show our approach has better effects than previous methods on both overlapping and non- overlapping communities.

Incorporating profile information in community detection for online social networks

Community structure is an important feature in the study of complex networks. It is because nodes of the same community may have similar properties. In this paper we extend two popular community detection methods to partition online social networks. In our extended methods, the profile information of users is used for partitioning. We apply the extended methods in several sample networks of Facebook. Compared with the original methods, the community structures we obtain have higher modularity. Our results indicate that users’ profile information is consistent with the community structure of their friendship network to some extent. To the best of our knowledge, this paper is the first to discuss how profile information can be used to improve community detection in online social networks.

대규모 네트워크에서 Modularity를 이용한 향상된 커뮤니티 추출 알고리즘

In this paper, an improved community detection algorithm based on the modularity is proposed. The existing algorithm does not consider the information that the nodes have in checking the possible modularity increase, hence the computation may be inefficient. The proposed algorithm computes the node degree (weight) and sorts them in non-increasing order. By checking the possible modularity value increase for the nodes in the nonincreasing order of node weights, the algorithm finds the final solution more quickly than the existing algorithm does. Through the computational experiments, it is shown that the proposed algorithm finds a modularity as good as the existing algorithm obtains.

A framework for joint community detection across multiple related networks

Community detection in networks is an active area of research with many practical applications. However, most of the early work in this area has focused on partitioning a single network or a bipartite graph into clusters/communities. With the rapid proliferation of online social media, it has become increasingly common for web users to have noticeable presence across multiple web sites. This raises the question whether it is possible to combine information from several networks to improve community detection. In this paper, we present a framework that identifies communities simultaneously across different networks and learns the correspondences between them. The framework is applicable to networks generated from multiple web sites as well as to those derived from heterogeneous nodes of the same web site. It also allows the incorporation of prior information about the potential relationships between the communities in different networks. Extensive experiments have been performed on both synthetic and real-life data sets to evaluate the effectiveness of our framework. Our results show superior performance of simultaneous community detection over three alternative methods, including normalized cut and matrix factorization on a single network or a bipartite graph.

Improving community detection in networks by targeted node removal

How a network breaks up into subnetworks or communities is of wide interest. Here we show that vertices connected to many other vertices across a network can disturb the community structures of otherwise ordered networks, introducing noise. We investigate strategies to identify and remove noisy vertices ("violators") and develop a quantitative approach using statistical breakpoints to identify when the largest enhancement to a modularity measure is achieved. We show that removing nodes thus identified reduces noise in detected community structures for a range of different types of real networks in software systems and in biological systems.