Algorithm For Community Detection In Networks Research Articles

A weighted prior tensor train decomposition method for community detection in multi-layer networks

Community detection in multi-layer networks stands as a prominent subject within network analysis research. However, the majority of existing techniques for identifying communities encounter two primary constraints: they lack suitability for high-dimensional data within multi-layer networks and fail to fully leverage additional auxiliary information among communities to enhance detection accuracy. To address these limitations, a novel approach named weighted prior tensor training decomposition (WPTTD) is proposed for multi-layer network community detection. Specifically, the WPTTD method harnesses the tensor feature optimization techniques to effectively manage high-dimensional data in multi-layer networks. Additionally, it employs a weighted flattened network to construct prior information for each dimension of the multi-layer network, thereby continuously exploring inter-community connections. To preserve the cohesive structure of communities and to harness comprehensive information within the multi-layer network for more effective community detection, the common community manifold learning (CCML) is integrated into the WPTTD framework for enhancing the performance. Experimental evaluations conducted on both artificial and real-world networks have verified that this algorithm outperforms several mainstream multi-layer network community detection algorithms.

Comparing the Clique Percolation algorithm to other overlapping community detection algorithms in psychological networks: A Monte Carlo simulation study.

In psychological networks, one limitation of the most used community detection algorithms is that they can only assign each node (symptom) to a unique community, without being able to identify overlapping symptoms. The clique percolation (CP) is an algorithm that identifies overlapping symptoms but its performance has not been evaluated in psychological networks. In this study, we compare the CP with model parameters chosen based on fuzzy modularity (CPMod) with two other alternatives, the ratio of the two largest communities (CPRat), and entropy (CPEnt). We evaluate their performance to: (1) identify the correct number of latent factors (i.e., communities); and (2) identify the observed variables with substantive (and equally sized) cross-loadings (i.e., overlapping symptoms). We carried out simulations under 972 conditions (3x2x2x3x3x3x3): (1) data categories (continuous, polytomous and dichotomous); (2) number of factors (two and four); (3) number of observed variables per factor (four and eight); (4) factor correlations (0.0, 0.5, and 0.7); (5) size of primary factor loadings (0.40, 0.55, and 0.70); (6) proportion of observed variables with substantive cross-loadings (0.0%, 12.5%, and 25.0%); and (7) sample size (300, 500, and 1000). Performance was evaluated through the Omega index, Mean Bias Error (MBE), Mean Absolute Error (MAE), sensitivity, specificity, and mean number of isolated nodes. We also evaluated two other methods, Exploratory Factor Analysis and the Walktrap algorithm modified to consider overlap (EFA-Ov and Walk-Ov, respectively). The Walk-Ov displayed the best performance across most conditions and is the recommended option to identify communities with overlapping symptoms in psychological networks.

Quantitative and qualitative interpretation of community partitions by map overlaying and calculating the distribution of related geographical features

Abstract Applying community detection algorithms in spatial interaction networks constructed from modern human communication records is an essential means of evaluating urban territorial subdivisions. Previous studies have usually involved qualitative rather than quantitative interpretations of community detection results. This article proposes a method of quantitatively and qualitatively interpreting community partition results by map overlaying the spatial regions corresponding to the detected communities with the related geographical features and by calculating the distribution of the geographical features contained in the regions and the entropy value of each distribution. The interpretation of the communities detected from the spatial interaction networks is carried out from the perspective of multi-temporal and multi-spatial scales and multi-geographical features. Extensive experiments were conducted with Milan, Italy, as the study area. The spatial interaction records reflected by telephone calls, land use, and point of interest (POI) data were used as the experimental data. Experimental results demonstrated the effectiveness of our method, and the specific results include: (1) Qualitative interpretation of multi-spatial resolution scale communities detected from the long-term aggregated spatial interaction network. The cohesiveness, homogeneity, and heterogeneity of the detected communities were qualitatively interpreted by the spatial distribution patterns of the land use dataset and the POI dataset. (2) Quantitative interpretation of multi-spatial resolution scale communities detected from the long-term aggregated spatial interaction network. The low spatial resolution scale community partitions and the high spatial resolution scale community partitions were interpreted through the statistical distribution of the land use dataset and the POI dataset, respectively. (3) Qualitative interpretation of the stable and active regions discovered from the community time series. Regardless of the community partitions’ spatial resolution scales, the stable and active regions were distinguished with the statistical distributions of the land use dataset and the POI dataset.

Dynamic community detection based on graph convolutional networks and contrastive learning

With the continuous development of technology and networks, real-life interactions are gradually being abstracted into social networks for study. Social circles are a fundamental structural feature that is prevalent on social networks. Thus, exploring social circle structure plays an important role in revealing the characteristics of complex social networks and provides guidance for understanding social behavior in real life. For example, it can aid in precision marketing, personalized recommendation, and knowledge dissemination within social circles. One of the important means of identifying social network circles lies on community detection algorithms. However, real-world social networks are often dynamic and can be studied and analyzed by building dynamic networks, while existing dynamic network community detection methods tend to ignore the global structure information and time-series information of nodes. To address this problem, this paper proposes a dynamic network community detection algorithm based on graph convolutional neural networks and contrastive learning, which fully captures the adjacent characteristics between nodes based on the correlation information and leverages the feature smoothing strategy to efficiently extract node representations of dynamic networks under unsupervised scenario. Specifically, the proposed algorithm first utilizes node correlation based aggregation strategy to compute the feature matrix for single time-step of the dynamic network. Then, mutual information maximization is implemented based on cross-entropy between learned local and global representations. To reduce the computational overhead in the optimization process, an additional LSTM module is further equipped for updating the parameters of graph convolutional networks in each time-step. Additionally, a contrastive learning based network smoothing strategy is designed to minimize the feature differences between neighboring nodes. Comparative experiments demonstrate that the proposed algorithm achieves excellent performance on both synthetic and real networks.

Evaluating Community Detection Algorithms for Multilayer Networks: Effectiveness of Link Weights and Link Direction

Analyzing the structures of multilayer networks (MLNs) has been a hot research topic in network science. Community detection algorithms are important tools for analyzing MLNs. In the literature, several community detection algorithms for MLNs have been proposed. Moreover, there are several options for the graph representation of an MLN: for example, directed or undirected, weighted or unweighted, and using information from all or only some layers. Although these options may affect the results of community detection in MLNs, representations that are effective for community detection have not yet been clarified. In this paper, we experimentally evaluate the effectiveness of three types of community detection algorithms for MLNs and examine how the graph representation of an MLN affects the results of these algorithms. Our main findings are as follows: (1) The flattening approach is particularly effective, whereas the layer-by-layer approach is not applicable to detecting communities in MLNs of Twitter users. (2) Using a directed graph for each layer of an MLN increases the accuracy of community detection. (3) The Leiden method, which is a community detection algorithm for single-layer networks, achieves comparable accuracy with the community detection algorithms for MLNs, which suggests that there exists room for improvement in multilayer community detection algorithms for effectively utilizing the multilayer structures of MLNs.

Network connectedness and the contagion structure of informed trading: Evidence from the time and frequency domains

In this paper, the network connectedness of informed trading between industries and its contagion structure are studied in the time and frequency domains based on Chinese stock markets. Using the spillover index method, we find significant interindustry spillover effects of informed trading. Specifically, Industrials, Consumer Discretionary, and Materials are the primary spillover sources that have strong outward spillover effects on all other industries, while Communication Services and Financial represent the top recipients in the market. In addition, using the network community detection algorithm, a 2-community structure of contagion is further revealed. The two communities are driven by Industrials and Consumer Discretionary, respectively. More important, the strong outward effects of Industrials, Consumer Discretionary, and Materials and their interlinkages would cause both intracommunity and intercommunity spillovers and result in market-wide informed trading contagion. Furthermore, the frequency decomposition analysis illustrates that interindustry spillovers of informed trading are usually dominated by low-frequency (6–200 days) connectedness. Finally, the contributing factor analysis demonstrates that macroeconomic changes are the main determinants of informed trading spillovers.

Reaction network simplification and key routes extraction for steam cracking process

The reaction mechanism is critical to understanding complex reaction systems in specific production processes in industries such as the petrochemical industry. In this study, we proposed a reaction network flow analysis/community-based mechanism detection/target species selection integrated framework to simplify the reaction network and extract key reaction routes. The naphtha steam cracking reaction system was preliminarily selected as an industrial example. The auto-generated reaction network was simultaneously set as an input of the PageRank-based reaction network analysis tool and community detection algorithm based on modularity analysis. After recording the importance and community affiliation information of each node in the network, the simplified communities were obtained and considered as skeletal mechanisms divided by reaction categories. Steps of target species selection were then implemented to search for the key nodes and routes in these skeletal mechanisms. The accuracy and trustworthiness of the results obtained by the framework were confirmed through a literature comparison and validation in an industrial example. The output results, including the simplified reaction networks, detected communities, and key reaction routes, are expected to bridge the gap between microscale information and macroscale phenomena.

A distributed community detection algorithm for large scale networks under stochastic block models

Community detection for large scale networks is of great importance in modern data analysis. In this work, we develop a distributed spectral clustering algorithm to handle this task. Specifically, we distribute a certain number of pilot network nodes on the master server and the others on worker servers. A spectral clustering algorithm is first conducted on the master to select pseudo centers. Next, the indexes of the pseudo centers are broadcasted to workers to complete the distributed community detection task using an SVD (singular value decomposition) type algorithm. The proposed distributed algorithm has three advantages. First, the communication cost is low, since only the indexes of pseudo centers are communicated. Second, no further iterative algorithm is needed on workers while a “one-shot” computation suffices. Third, both the computational complexity and the storage requirements are much lower compared to using the whole adjacency matrix. We develop a Python package DCD (The Python package is provided in https://github.com/Ikerlz/dcd.) to implement the distributed algorithm on a Spark system and establish theoretical properties with respect to the estimation accuracy and mis-clustering rates under the stochastic block model. Experiments on a variety of synthetic and empirical datasets are carried out to further illustrate the advantages of the methodology.

Comparing community detection algorithms in psychometric networks: A Monte Carlo simulation.

Identifying the correct number of factors in multivariate data is fundamental to psychological measurement. Factor analysis has a long tradition in the field, but it has been challenged recently by exploratory graph analysis (EGA), an approach based on network psychometrics. EGA first estimates a network and then applies the Walktrap community detection algorithm. Simulation studies have demonstrated that EGA has comparable or better accuracy for recovering the same number of communities as there are factors in the simulated data than factor analytic methods. Despite EGA's effectiveness, there has yet to be an investigation into whether other sparsity induction methods or community detection algorithms could achieve equivalent or better performance. Furthermore, unidimensional structures are fundamental to psychological measurement yet they have been sparsely studied in simulations using community detection algorithms. In the present study, we performed a Monte Carlo simulation using the zero-order correlation matrix, GLASSO, and two variants of a non-regularized partial correlation sparsity induction methods with several community detection algorithms. We examined the performance of these method-algorithm combinations in both continuous and polytomous data across a variety of conditions. The results indicate that the Fast-greedy, Louvain, and Walktrap algorithms paired with the GLASSO method were consistently among the most accurate and least-biased overall.

Community Detection Algorithms in Healthcare Applications: A Systematic Review

Over the past few years, the number and volume of data sources in healthcare databases has grown exponentially. Analyzing these voluminous medical data is both opportunity and challenge for knowledge discovery in health informatics. In the last decade, social network analysis techniques and community detection algorithms are being used more and more in scientific fields, including healthcare and medicine. While community detection algorithms have been widely used for social network analysis, a comprehensive review of its applications for healthcare in a way to benefit both health practitioners and the health informatics community is still overwhelmingly missing. This paper contributes to fill in this gap and provide a comprehensive and up-to-date literature research. Especially, categorizations of existing community detection algorithms are presented and discussed. Moreover, most applications of social network analysis and community detection algorithms in healthcare are reviewed and categorized. Finally, publicly available healthcare datasets, key challenges, and knowledge gaps in the field are studied and reviewed.

Innovation evolution of industry-university-research cooperation under low-carbon development background: In case of 2 carbon neutrality technologies

Offsetting carbon emissions by various means is carbon neutrality. Technological innovation is an important path to carbon neutrality and has been framed in China’s policy planning, such as innovations about cleaner combustion technologies. To understand the current innovation of clean technology in gas boilers, this paper selects 2 Carbon Neutrality Technologies (CNT) identified in China, respectively low nitrogen combustion and catalytic combustion, to obtain relevant patent data of innovative technology. Besides, social network analysis (SNA) and community detection algorithms are used to show the evolution of three stages from 2005 to 2020. On the whole, several enterprises which dominate the network in the innovation cooperation of clean combustion technology in China exist at all stages, and the participation consciousness of scientific research institutions in universities is not strong. Network analysis shows that the network average degree fluctuates (1.172–1.962–1.442) with no obvious changes, and the network becomes sparse gradually. The analysis of the two-mode network shows that B01J refers to the most technical field involved in the cooperation except the first stage. B01D, F23G, and F23D are the key areas of technological innovation in the three stages except for B01J. The current innovations in clean combustion technology are not so much about combustion itself, instead, they focus on general physical and chemical methods. But at Stage2 and Stage3, more attention has begun to shift to burner innovation.

Know thy tools! Limits of popular algorithms used for topic reconstruction

Abstract To reconstruct topics in bibliometric networks, one must use algorithms. Specifically, researchers often apply algorithms from the class of network community detection algorithms (such as the Louvain algorithm) that are general-purpose algorithms not intentionally programmed for a bibliometric task. Each algorithm has specific properties “inscribed,” which distinguish it from the others. It can thus be assumed that different algorithms are more or less suitable for a given bibliometric task. However, the suitability of a specific algorithm when it is applied for topic reconstruction is rarely reflected upon. Why choose this algorithm and not another? In this study, I assess the suitability of four community detection algorithms for topic reconstruction, by first deriving the properties of the phenomenon to be reconstructed—topics—and comparing if these match with the properties of the algorithms. The results suggest that the previous use of these algorithms for bibliometric purposes cannot be justified by their specific suitability for this task.

Multiplex network community detection algorithm based on motif awareness

Multiplex networks have shown high utility in modeling complex systems, e.g., academic social networks. Composite community structures contribute to revealing meaningful grouping patterns in multiplex networks. Multiplex network community detection, especially using higher-order interaction information, has become a popular research topic in complex network analysis. We propose a multiplex network community detection algorithm based on motif awareness (CDMA) to reduce information loss during the aggregation of multiplex networks, which can solve the community detection problem for cold start nodes and improve the quality of multiplex network community detection. Interlayer differences are fully considered in the topology of the multiplex network throughout the algorithm. First, the motif is used to mine the higher-order interaction information in each layer of the network topology, and a multiplex network aggregation model is constructed to aggregate the multiplex network into a single-layer composite network. In addition, the information propagation method is used to transform the node information in the single-layer composite network into a vector. Furthermore, we propose a node importance evaluation index and select the nodes with the greatest significance to solve the initial node selection sensitivity problem of the k-means algorithm. The comparative experimental results on real multiplex network datasets, e.g., SCHOLAT, show that the proposed CDMA algorithm can play an important role in addressing the issue of effective detection of the community structure of multiplex networks with higher accuracy.

Detecting Community Structure in Financial Markets Using the Bat Optimization Algorithm

A lucid representation of the hidden structure of real-world application has attracted complex network research communities and triggered a vast number of solutions in order to resolve complex network issues. In the same direction, initially, this paper proposes a methodology to act on the financial dataset and construct a stock correlation network of four stock indexes based on the closing stock price. The significance of this research work is to form an effective stock community based on their complex price pattern dependencies (i.e., simultaneous fluctuations in stock prices of companies in a time series data). This paper proposes a community detection approach for stock correlation complex networks using the BAT optimization algorithm aiming to achieve high modularity and better-correlated communities. Theoretical analysis and empirical modularity performance measure results have shown that the usage of BAT algorithm for community detection proves to transcend performance in comparison to standard network community detection algorithms – greedy and label propagation.

Multiscale analysis of cancer service areas in the United States

The purpose of delineating Cancer Service Areas (CSAs) is to define a reliable unit of analysis, more meaningful than geopolitical units such as states and counties, for examining geographic variations of the cancer care markets using geographic information systems (GIS). This study aims to provide a multiscale analysis of the U.S. cancer care markets based on the 2014–2015 Medicare claims of cancer-directed surgery, chemotherapy, and radiation. The CSAs are delineated by a scale-flexible network community detection algorithm automated in GIS so that the patient flows are maximized within CSAs and minimized between them. The multiscale CSAs include those comparable in size to those 4 census regions, 9 divisions, 50 states, and also 39 global optimal CSAs that generates the highest modularity value. The CSAs are more effective in capturing the U.S. cancer care markets because of its higher localization index, lower cross-border utilizations, and shorter travel time. The first two comparisons reveal that only a few regions or divisions are representative of the underlying cancer care markets. The last two comparisons find that among the 39 CSAs, 54% CSAs comprise multiple states anchored by cities near inner state borders, 28% are single-state CSAs, and 18% are sub-state CSAs. Their (in)consistencies across state borders or within each state shed new light on where the intervention of cancer care delivery or the adjustment of cancer care costs are needed to meet the challenges in the U.S. cancer care system. The findings could guide stakeholders to target public health policies for more effective coordination of cancer care in improving outcomes and reducing unnecessary costs.

A subgraphs-density based overlapping community detection algorithm for large-scale complex networks

A subgraphs-density based overlapping community detection algorithm for large-scale complex networks

Community Detection in Semantic Networks: A Multi-View Approach.

The semantic social network is a complex system composed of nodes, links, and documents. Traditional semantic social network community detection algorithms only analyze network data from a single view, and there is no effective representation of semantic features at diverse levels of granularity. This paper proposes a multi-view integration method for community detection in semantic social network. We develop a data feature matrix based on node similarity and extract semantic features from the views of word frequency, keyword, and topic, respectively. To maximize the mutual information of each view, we use the robustness of L21-norm and F-norm to construct an adaptive loss function. On this foundation, we construct an optimization expression to generate the unified graph matrix and output the community structure with multiple views. Experiments on real social networks and benchmark datasets reveal that in semantic information analysis, multi-view is considerably better than single-view, and the performance of multi-view community detection outperforms traditional methods and multi-view clustering algorithms.

A complex network community detection algorithm based on random walk and label propagation

AbstractThe community structure is proving to have a very important role in the understanding of complex networks, but discovering them remains a very difficult problem despite the existence of several methods. In this article, we propose a novel algorithm for discovering communities in complex networks based on a modified random walk (RW) and label propagation algorithm (LPA). First, we calculate the similarity between nodes based on the new formula of RW. Then, the labels are propagated by the obtained similarity of the first step using LPA. Finally, the third step will be a new measure to find the optimal partitioning of communities. Experimental results obtained on several real and synthetic networks reveal that our algorithm outperforms existing methods in finding communities.

Network analysis of comorbidity patterns in heart failure patients using administrative data

Background: Congestive Heart Failure (HF) is a widespread chronic disease characterized by a very high incidence in elder people. The high mortality and readmission rate of HF strongly depends on the complicated morbidity scenario often characterising it.
 Methods: Data were retrieved from the healthcare administrative datawarehouse of Lombardy, the most populated regional district in Italy. Network analysis techniques and community detection algorithms are applied to comorbidities registered in hospital discharge papers of HF patients, in 7 cohorts between 2006 and 2012.Results: The relevance network indexes applied to the 7 cohorts identified death, ipertension, arrythmia, renal and pulmonary diseases as the most relevant nodes related to HF, in terms of prevalence and closeness/strenght of the relationship. Moreover, 3 clusters of nodes have been identified in all the cohorts, i.e. those related to cancer, lung diseases and heart/circulation related problems.Conclusions: Network analysis can be a useful tool in epidemiologic framework when relational data are the objective of the investigation, since it allows to visualize and make inference on patterns of association among nodes (here HF comorbidities) by means of both qualitative indexes and clustering techniques.

A new metric to compare local community detection algorithms in social networks using geodesic distance

A new metric to compare local community detection algorithms in social networks using geodesic distance