Search

With the rapid development of internet technologies and the increasing availability of large-scale data, the detection of community structures within complex networks has become a critical area of research. This paper introduces a novel community detection technique called LB-SAM (Local Beam Search with Simulated Annealing and Modularity), designed to efficiently uncover hidden community structures in large-scale social networks. LB-SAM integrates Local Beam Search (LBS) to explore the local network structure and Simulated Annealing (SA) to globally optimize modularity, enabling the detection of communities with intricate boundaries and strong internal connections. By focusing on influential nodes to form subgroups and recursively merging them based on modularity, LB-SAM provides superior scalability and robustness in both real-world and synthetic networks. Extensive experiments conducted on 12 real-world and 6 synthetic datasets demonstrate that LB-SAM consistently outperforms existing state-of-the-art algorithms, particularly in networks with unclear community structures, and scales effectively to billion-scale networks. The proposed method has wide-ranging applications in sociology, biology, marketing, and cybersecurity, offering valuable insights into the structure and dynamics of large social networks.

Recommendation systems play an important role in creating personalized content for consumers, improving their overall experiences across several applications. Providing the user with accurate recommendations based on their interests is the recommender system’s primary goal. Collaborative filtering-based recommendations with the help of matrix factorization techniques is very useful in practical uses. Owing to the expanding size of the dataset and as the complexity increases, there arises an issue in delivering accurate recommendations to the users. The efficient functioning of the recommendation system undergoes the scalability challenge in controlling large and varying datasets. This paper introduces an innovative approach by integrating matrix factorization techniques and community detection methods where the scalability in recommendation systems will be addressed. The steps involved in the proposed approach are: 1) The rating matrix is modeled as a bipartite network. 2) Communities are generated from the network. 3) Extract the rating matrices that belong to the communities and apply MF to these matrices in parallel. 4) Merge the predicted rating matrices belonging to the communities and evaluate root mean square error (RMSE), mean square error (MSE), and mean absolute error (MAE). In our paper different matrix factorization approaches like basic MF, NMF, SVD++, and FANMF are taken along with the Louvain community detection method for dividing the communities. The experimental analysis is performed on five different diverse datasets to enhance the quality of the recommendation. To determine the method’s efficiency, the evaluation metrics RMSE, MSE, and MAE are used, and the time required to evaluate the computation is also computed. It is observed in the results that almost 95% of our results are proven effective by getting lower RMSE, MSE, and MAE values. Thus, the main aim of the user will be satisfied in getting accurate recommendations based on the user experiences.

Administrative divisions are regional divisions of the state for the purpose of hierarchical administration. With the development of urbanization process, a concern arising from the changing urban dynamics is that whether current administrative division accords with urban development? Following the concept of urban space formed by human activities, existing studies have sought proxies for human activities and then delineated urban boundaries formed by human activities from a data-driven perspective. However, their lack of exploration into the implicit human activity patterns behind the proxies results in weak interpretability and rationality of methods and results. This paper studies the delineation of urban interior boundaries respecting to human mobility patterns. Specifically, this paper first uses human mobility as a proxy to construct a network that represents spatial interaction in urban regions. Then we explore the region-based human crowd mobility patterns to reveal the mechanisms behind the formation of urban space through human activities. Finally, we employ the community detection technology to naturally delimit the urban interior boundaries formed by human mobility, and make a comparison with the official urban boundaries. Taking Xi’an in Shaanxi Province of China as an example, we conduct the delineation on the spatial interaction network based on the real mobility information of 24,770,715 mobile phone users in the whole city. We find that human mobility can establish a stable correlation between regions (or capture the objective correlations between regions), and the human crowd patterns are applicable for mining unusual urban regions from the perspective of anomaly detection, which are of great significance for understanding the urban spatial structure based on empirical evidence. In addition, in the final delineation result, some unexpected communities that are closely linked due to human activities appear from the results, and these findings help the urban planners re-examine the administrative division.

In this study, we propose the User2Vec framework, a novel approach for capturing user behavior based on the co-occurrence of user activities. User behavior modeling is essential for understanding relationships in areas such as e-learning and marketing, but most existing methods rely on explicit social interaction data, which are not always available. We propose the User2Vec framework, inspired by the Word2Vec model, to encode user activity into dense vector representations using the temporal co-occurrence of active users. This approach allows the identification of user groups without relying on direct interaction data. The proposed methodology involves preprocessing user activity logs, extracting temporal co-occurrence features, and applying clustering techniques such as k-means and spectral clustering to reveal latent structures. These clustering methods were chosen for their balance of computational efficiency and effectiveness in capturing implicit relationships. The approach is experimented on a dataset from a university e-learning platform, where student logs are processed to cluster users according to their co-occurring actions. Comparative evaluations show that the proposed model outperforms previous co-occurrence-based methods in terms of clustering accuracy and computational efficiency. Our results highlight the significant potential of the proposed framework in applications such as personalized learning and community detection.

Community detection is a very important research topic in the field of Social Network Analysis. Lots of researchers are working in this field due to its applications in various fields like medicine, social, Business, Marketing, and research. Researchers are proposing new algorithms to detect the communities having better performance as compared to the existing techniques. Initially, Newman and Girvan proposed traditional algorithms for community detection from social networks in 2004, but with the growth of social networks, Convolutional Neural Network (CNN) based algorithms are proposed by different researchers in recent years due to the inefficiency of traditional methods. After reviewing the state-of-the-art algorithms based on CNN learned that MRFasGCN is having the best performance compared to any other state-of-the-art algorithms for large data sets. In this algorithm, researchers have integrated the technique of Graph Convolutional Neural Network (GCN) with the statistical model Markov Random Field (MRF) to get better results and after implementing it on large datasets comparison is done on its results with other state-of-the-art algorithms and got to know its performance is far better than any other algorithm. While MRFasGCN is performing well on social networks and provides ground truth communities, there is a possibility available for improvement due to the sparsity problem. This paper proposes a new algorithm called Modified MRFasGCN. Two modifications are done to the existing algorithm, 1. In pre-processing, rather than passing the adjacency matrix with the normalized adjacency matrix, it will pass the reconstructed adjacency matrix and normalized reconstructed adjacency matrix, which resolves the sparsity problem, 2. GCN layer output will be fed to the MRF layer and refined results will be passed to the Adam optimizer without subtraction. Our experimental analysis shows that the modified algorithm provides better ground truth communities than the MRFasGCN and solves the problem of sparsity as passes a reconstructed adjacency matrix. In this paper, the proposed algorithm is executed on different datasets having different sizes like CORA (2708 Nodes), Flicker (80513 Nodes) and DBLP (317080 Nodes) and compared on different Community Detection metrics like accuracy, NMI, F1 Score, and execution time with other algorithms. After Experiments NMI value for MRFasGCN on DBLP data set is 0.662 while for Modified MRFasGCN it is 0.672, Modified MRFasGCN algorithm provides significant improvement of 2.9% in performance as compared to MRFasGCN. F1-Score of proposed algorithm is 0.511 on DBLP dataset which is better than MRFasGCN.

Community discovery is an essential research area with significant real-world applications. Lately, Graph Convolutional Networks (GCNs) have gained popularity for their ability to effectively identify and represent node relationships in low-dimensional spaces, enhancing community detection. However, previous studies have ignored two problems: insufficient network node labeling and incomplete information capture of similar dependencies between nodes, which affect the understanding and analysis of the network and the accuracy of node embedding. To this end, this paper proposed GCN-based encoder-decoder networks with Markov random field for unsupervised community detection. Specifically, we propose a comprehensive framework for enhancing the representation of learning nodes, which consists of the following components: a GCN-based encoder-decoder module, an MRF-based information enhancement module, and a modularity maximization module. The GCN-based encoder-decoder module uses graph attention autoencoder and dual decoder to learn node embedding matrix, which solves the problems of insufficient node labels and unclear information. The MRF-based information enhancement module uses the Markov random field to comprehensively consider the characteristics of the node itself and the relationship between the nodes to improve the representation ability of node embedding and solves the problem of incomplete information acquisition of similar dependencies between nodes. The modularity maximization module divides the network into subgroups with tight connections and divides similar nodes into the same community. Finally, we conducted node clustering experiments on three datasets, which showed improvements of approximately 0.3 percent, 3 percent, and 5 percent compared to the state-of-the-art baseline, respectively.

In this study, a novel hybrid recommendation model is designed for popular social media platforms such as Facebook, Twitter, WhatsApp, and Instagram. Recognizing the global significance of these platforms, the model aims to facilitate connections among users with similar interests, leveraging community detection algorithms. By integrating sentimental analysis, link prediction, and community detection techniques, a comprehensive network is constructed that enhances user experience and addresses the escalating issue of cyberbullying in the era of increasing connectivity. The proposed model strategically combines nodes and edges to form communities tailored to individual user interests. Social media comments, gathered randomly from the Facebook API, undergo sentimental analysis to evaluate the model’s performance. The innovative Community-guided Link Prediction algorithm using Sentimental Analysis (CLPSA) for multilayer networks is employed, aiding in the identification of nodes associated with fake comments and thereby mitigating cyberbullying problems. Integrating sentiment analysis enhances understanding of human emotions, and users in the same cluster are more likely to connect in the future, promoting positive engagement. The model’s novelty lies in its utilization of the Proposed Cluster Edge Betweenness algorithm, which identifies clusters of users with denser connections among them while being loosely connected to other separate modules. This approach helps create a network structure that fosters positive interactions and reduces the occurrence of intolerable comments. The experimental comparison of community detection algorithms underscores the effectiveness of our approach, particularly the superior performance of the Cluster Edge Betweenness algorithm. It achieves high accuracy in identifying user interests and forming cohesive clusters, essential for enhancing user engagement and community cohesion within social media networks. The qualitative analysis of comments leads to the formation of communities within the network, uniting users with similar interests into cohesive clusters. The uniqueness of the approach lies in the integration of sentimental analysis, link prediction, and community detection algorithms within a multilayer network framework. Specifically, the proposed Cluster Edge Betweenness algorithm is compared with two other community detection algorithms (Cluster Label Proportionate and ClusterFastGreedy). Results demonstrate the superior accuracy of the algorithm, achieving a 95% accuracy in aligning with user interests and a notably high modularity density of 0.2331, indicative of superior clustering quality. By incorporating these elements, the model can evolve into a more comprehensive solution for addressing cyberbullying and promoting positive engagement in social media networks.

Social IoT (SIoT) refers to socializing in the Internet of Things (IoT), where things generate social relationships. Due to the development of objects and issues such as delayed response, slow search, and composite service process, distributed object service discovery, selection, and composition based on the social structure have become essential challenges in the SIoT. Therefore, it is necessary to provide an efficient method for evaluating the effectiveness of service discovery in identifying suitable devices to offer requested services and the best composition strategy for combining requested services. This paper presents a new community detection algorithm that detects IoT devices with social connections in the SIoT network to facilitate service discovery and composition by reducing search space. Additionally, it introduced a new service provisioning algorithm to optimize service discovery and composition, called An Improved Genetic Algorithm based on Community Detection (IGA-CD). Its effectiveness in detected communities is better than other methods in computation modularity, execution time, and cluster assignment quality methods by determining the network’s ideal devices. Experimental results demonstrate the efficacy of the proposed algorithms, which outperform other approaches in terms of scalability, efficiency, and flexibility. The IGA-CD average execution time is 0.129 seconds, which proves its efficiency and faster composition.

Community detection is an important task in complex network analysis. A community is a set of cohesive vertices that have more connections within the set than outside. In many real Complex Networks (CNs), these communities naturally overlap, meaning an individual node can belong to more than one community. This overlapping structure is crucial for many real applications, such as social influence detection, cyberattack detection, and recommendation systems. Existing methods often struggle to capture both network topology and node features, leading to suboptimal overlapping community detection. In this paper, we propose an efficient method called GCNFCM, which utilizes Graph Convolutional Networks (GCNs), Fuzzy C-means (FCM), and the modularity Q algorithm for overlapping community detection. The key idea is to achieve robust feature learning for nodes and then identify the best structure for overlapping community detection. GCNFCM extracts node embeddings from CNs, considering both topology and attributes through a dual-decoder design (inner product and GCN), while FCM is employed for optimal overlapping community detection. Furthermore, FCM is guided by the modularity Q algorithm for accurate community identification without requiring prior knowledge of the community count. Experimental results on ten real-world CNs of varying sizes demonstrate that our proposed method outperforms other state-of-the-art overlapping community detection methods in terms of producing cohesive communities and identifying ground-truth communities. Additionally, the results indicate that the developed method effectively identifies good overlapping communities in real-world networks.

Financial distress prediction is crucial to financial planning, particularly amid emerging uncertainties. This study introduces a novel methodology for predicting financial distress by amalgamating network analysis and machine learning techniques. The approach involves establishing two company networks based on their similarity and correlation in crucial financial indicators. The first network reflects similarity across five features, while the second captures correlation in the most critical feature. Subsequently, seven network-centric features are extracted and integrated into the dataset as new variables. Community detection algorithms are also applied to cluster companies, with the resulting labels added as categorical variables. This process yields a modified dataset comprising both initial and network-based variables. Five classification algorithms are employed to forecast financial distress across three scenarios. Initially, models are trained using only the initial features. In subsequent scenarios, network-centric features from similarity and correlation networks are incorporated, enhancing the predictive accuracy of machine learning models. Notably, features from the similarity network play a pivotal role in this improvement. The proposed model showcases superior predictive capabilities and offers a holistic understanding of the dynamic interactions among financial entities. The results underscore the efficacy of network-based strategies in refining financial distress prediction models, providing valuable insights for decision-makers.

Given the nature of time series and their vast applications, it is essential to find clustering algorithms that depict their real-life properties. Among the features that can hugely effect the options available for time series are overlapping and hierarchical properties. In this paper a novel approach to analyze time series with such features is introduced. Using the two concepts of network construction and link community detection, we have attempted to analyze and identify the mentioned properties of time series using data that is often gathered first hand. The proposed algorithm has been applied using both recent and common similarity measures on ten synthetic time series with hierarchal and overlapping features, alongside various distance measures. When testing the proposed approach, the element-centric measure of similarity indicated a clear increased accuracy for this algorithm, showing the highest accuracy when used alongside the Dynamic Time Warping distance measure. Moreover, the proposed algorithm has been very successful in identifying and forming communities for both large and small time series, thus solving another one of the main issues previous algorithms tended to have.

Expert profiling aims to discover the expertise of an author. This task is useful for identifying the research groups existing within an organization as well as measuring the similarities between authors’ expertise. Thus, identifying areas of expertise becomes a critical part of this task, especially in cases where the publications are unannotated. Commonly used topic modeling methods such as Latent Dirichlet Allocation still fall short in determining the number of topics automatically and discovering the hierarchical relationships between topics. To solve these issues, we adopted a graph-based approach which constructs a graph from publication features such as authors and keywords (Silva et al., 2018). We applied the Louvain algorithm repeatedly to discover the topics with hierarchical order automatically. We utilize keyword extraction methods to generate keywords for each respective publication to handle the missing values. We perform experiments to determine the optimum HPMI value. Results showed that graphs constructed from default and SIFRank keywords with transformation weights of $\alpha =0.5$ and $\beta =1.0$ produce topics with the best HPMI score. We evaluate the profiles from this method (CDT) with ATM as the baseline. It is shown that CDT produces better MAP, MRR, and nDCG scores than ATM. The work in this manuscript shows how community detection and keyword extraction could be utilized in expert profiling tasks. Our observation shows that the Louvain algorithm used only cluster publications into one topic, and thus still has limitations in classifying multidisciplinary publications. Further development could be done to handle such publications and increase the quality of keywords.

Heterogeneous information networks provide abundant structural and semantic information. Two main strategies for leveraging this data include meta-path-based and meta-path-free methods. The effectiveness of the former heavily depends on the quality of manually defined meta-paths, which may lead to the instability of the model. However, the existing meta-path-free methods lack of neighbor screening during aggregating, and there is also an overemphasis on attribute information. To address these issues, we propose the Heterogeneous Graph Neural Network model by incorporating Quantitative Sampling and Structure-aware Attention. We introduce a Quantitative Sampling Module that calculates the similarity between neighbors of the target nodes and target nodes, enabling us to select the top k nodes with the strongest relevance to the target node based on this measure, and incorporate a Structure-aware Attention Module during the aggregation of neighbor information. This module combines both structural and attribute information to aggregate the neighbor information effectively. By implementing these improvements, our proposed model exhibits superior performance compared to several state-of-the-art methods on two real-world datasets.

Networks are commonly used to model complex systems. The different entities in the system are represented by nodes of the network and their interactions by edges. In most real life systems, the different entities may interact in different ways necessitating the use of multiplex networks where multiple links are used to model the interactions. One of the major tools for inferring network topology is community detection. Although there are numerous works on community detection in single-layer networks, existing community detection methods for multiplex networks mostly learn a common community structure across layers and do not take the heterogeneity across layers into account. In this paper, we introduce a new multiplex community detection method that identifies communities that are common across layers as well as those that are unique to each layer. The proposed method, Multiplex Orthogonal Nonnegative Matrix Tri-Factorization, represents the adjacency matrix of each layer as the sum of two low-rank matrix factorizations corresponding to the common and private communities, respectively. Unlike most of the existing methods which require the number of communities to be pre-determined, the proposed method also introduces a two stage method to determine the number of common and private communities. The proposed algorithm is evaluated on synthetic and real multiplex networks, as well as for multiview clustering applications, and compared to state-of-the-art techniques.

Social networks are becoming increasingly popular and significant. One of the most distinctive features of these networks is their dynamic nature, which means that they change over time. Consequently, the community structure on these platforms also changes with time, making the detection of community structure a crucial area of research. Specifically, there is still a lack of understanding of how social networks and communities evolve over time. In this paper, we reveal that individual changing topics (i.e., individual temporal semantics) are a vital factor that drives community evolution. A novel dynamic community detection model is proposed, which takes into account natural evolutionary features. The model first partitions social networks into snapshots. It then detects the community structure at each snapshot by utilizing individual changing topics and information from the previous snapshot. Finally, the evolution of users’ interested topic distributions and topic distributions of communities are identified. The model is compared with five state-of-the-art baselines on two real datasets, and the experimental results demonstrate that our model outperforms all baselines.

Identifying key information from complex networks is of great practical significance for discovering the community structure. How to effectively use the information of node connections in the network and the attribute information in the attribute network is a major challenge in the current community detection problem of attribute networks. A graph convolutional network based on personalized web page ranking algorithm is proposed for community detection in attribute networks in this paper. First, the proposed algorithm uses the strong characteristics of graph convolution algorithm for integrating node topology and attribute information, and combines with personalized web page ranking algorithm to decouple the prediction process and propagation process in the model. In addition, the improved density peak detection method is used to sample the local structure center as a training set for training the algorithm model. Finally, k-means method is used to cluster the node vector representation results and get the community division. The experimental results on 7 datasets with 13 comparison algorithms show that the proposed algorithm has obvious improvement for attribute community detection.

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.

The detection of community structures is a very crucial research area. The problem of community detection has received considerable attention from a large portion of the scientific community. More importantly, these articles are spread across a large number of different disciplines, from computer science, to statistics, and social sciences. The analysis of modern social networks becomes rather cumbersome, as their size and number keeps growing larger and larger. Moreover, in the modern communities, users participate in large number of groups. From the network perspective, efficient methods should be developed to automatically identify overlapping communities, that is, communities with overlapping nodes. In this work, we use a probabilistic network model to characterize and identify linked communities with common nodes. The innovative idea in this work is that the communities are represented as Markovian networks with continuously changing states. Each state represents the number of users within a cluster, that have specific characteristic classes. Based on the current state, we introduce a fast, linear on the number of newly added users, approach to estimate the probability of each cluster to be homogeneous in terms of sets of user characteristics and to determine how well the new user fit within a community. Because of the linear computations involved, our proposed probabilistic model can detect communities and overlaps with low execution time and high accuracy, as shown in our experimental results. The experimental results have shown that our probabilistic scheme executes faster and provides more robust communities compared to competitive schemes.

Community detection is an extremely important task for complex network analysis. There still remains a challenge of how to improve the performance of community detection in real-world scenario. Some researchers think that the content in networks is helpful to identify communities, and also focus on combining network topology with node contents, alongside the eradication of inauspicious performance. Furthermore, network topology is often sparse, which is reflected in the lack of capability to represent communities. To address the above problems, this study identifies a novel non-negative matrix factorization method which both employs network embedding to enhance the representation power of network topology for communities and also integrates network topology and node contents for further raising the quality of community detection. Furthermore, we then obtain the parameters of the model for finding communities which is based on model inference. Alongside both synthetic and real-world networks with ground-truths, we compare the new method with the state-of-the-art methods. Experimental results show that the new method obtains significant improvement for community detection both by incorporating node contents and by enhancing network topology.

Social media networks have become a prime source for sharing news, opinions, and research accomplishments in various domains, and hundreds of millions of posts are announced daily. Given this wealth of information in social media, finding related announcements has become a relevant task, particularly in trending news (e.g., COVID-19 or lung cancer). To facilitate the search of connected posts, social networks enable users to annotate their posts, e.g., with hashtags in tweets. Albeit effective, an annotation-based search is limited because results will only include the posts that share the same annotations. This paper focuses on retrieving context-related posts based on a specific topic, and presents PINYON, a knowledge-driven framework, that retrieves associated posts effectively. PINYON implements a two-fold pipeline. First, it encodes, in a graph, a CORPUS of posts and an input post; posts are annotated with entities for existing knowledge graphs and connected based on the similarity of their entities. In a decoding phase, the encoded graph is used to discover communities of related posts. We cast this problem into the Vertex Coloring Problem, where communities of similar posts include the posts annotated with entities colored with the same colors. Built on results reported in the graph theory, PINYON implements the decoding phase guided by a heuristic-based method that determines relatedness among posts based on contextual knowledge, and efficiently groups the most similar posts in the same communities. PINYON is empirically evaluated on various datasets and compared with state-of-the-art implementations of the decoding phase. The quality of the generated communities is also analyzed based on multiple metrics. The observed outcomes indicate that PINYON accurately identifies semantically related posts in different contexts. Moreover, the reported results put in perspective the impact of known properties about the optimality of existing heuristics for vertex graph coloring and their implications on PINYON scalability.

The study of community detection in networks has drawn great attention in recent years. To find communities and to understand community semantics, both network topology and network content are utilized. Unfortunately, none of them can explain the driving factors of generating community structure with semantics, which is significant for understanding the mechanisms of community generation. Our observations on a large number of networks show that specific user social behaviors are underlying factors for the generation of community structure. We exploit four types of social behaviors that widely exist in networks, i.e., reciprocity of interactions, posting preference, multitopic preference, and temporal variation of topics. We investigate their impacts on the formation process of links and content in networks, during which communities with topics form. Our analysis shows that they are highly related to community structure. Consequently, a generative community detection model SBCD (social behavior-based community detection) is proposed by combining network topology and content, in which the above social behaviors play a core role. The model is evaluated on two real datasets. The experimental results show that SBCD outperforms state-of-the-art baselines. Finally, a case study illustrates several significant observations with respect to the proposed social behaviors.

Event detection is a popular research problem aiming to detect events from various data sources, such as news texts, social media postings or social interaction patterns. In this work, event detection is studied on social interaction and communication data via tracking changes in community structure and communication trends. With this aim, various community structure and communication trend based event detection methods are proposed. Additionally, a new strategy called community size range based change tracking is presented such that the proposed algorithms can focus on communities with different size ranges, and considerable time efficiency can be obtained. The event detection performance of the proposed methods is analyzed using a set of real world and benchmark data sets in comparison to previous solutions in the literature. The experiments show that the proposed methods have higher event detection accuracy than the baseline methods. Additionally, their scalability is presented through analysis by using high volume of communication data. Among the proposed methods, CN-NEW, which is a community structure based method, performs the best on the overall. The proposed communication trend based methods perform better mostly on communication data sets (such as CDR), whereas community structure based methods tend to perform better on social media-based data sets.

Community detection is a flourishing research field with a plethora of applications ranging from biology to sociology. Local community detection has emerged as a promising subfield of research concerned with community identification around a set of seeding nodes. The practical significance of local community detection is important for numerous real-world applications such as protein interactions and targeted advertising. Since 2005, when the first research paper on local community detection appeared, the literature has been vast and difficult to navigate, as each method works best under certain conditions and assumptions regarding the seed nodes and the identification of their community. For this reason, and motivated by the many real-world applications of local community detection, in this paper we provide a comprehensive overview and taxonomy of local community detection algorithms. There are quite a lot of surveys on community detection that make a compendious reference to local community detection. However, they do not achieve a systematic and comprehensive coverage of this particular field. Since the research area of local community detection is quite extensive, it is necessary to categorize and discuss the various methods, techniques, and assumptions used to address the problem. This survey aims to fill this gap and help researchers get a clear overview of the local community detection problem. To this end, we have also gathered the best documented tools and the most commonly used datasets in the local community detection literature to help researchers identify the tools they can use to prove their methods.

Modern data analysis and processing tasks typically involve large sets of structured data. Graphs provide a powerful tool to describe the structure of such data, where the entities and the relationships between them are modeled as the nodes and edges of the graph. Traditional single layer network models are insufficient for describing the multiple entity types and modes of interaction encountered in real-world applications. Recently, multi-layer network models, which consider the different types of interactions both within and across layers, have emerged to model these systems. One of the important tools in understanding the topology of these high-dimensional networks is community detection. In this paper, a joint nonnegative matrix factorization approach is proposed to detect the community structure in multi-layer networks. The proposed approach models the multi-layer network as the union of a multiplex and bipartite network and formulates community detection as a regularized optimization problem. This optimization problem simultaneously finds the nonnegative low-rank embedding of the intra- and inter-layer adjacency matrices while minimizing the distance between the two to guarantee pair-wise similarity across embeddings. The proposed approach can detect the community structure for both homogeneous and heterogeneous multi-layer networks and is robust to noise and sparsity. The performance of the proposed approach is evaluated for both simulated and real networks and compared to state-of-the-art methods.

Adopting Internet-of-things (IoT) in large-scale environments such as smart cities raises compatibility and trustworthiness challenges, hindering conventional service discovery and network navigability processes. The IoT network is known for its highly dynamic topology and frequently changing characteristics (e.g., the devices’ status, such as battery capacity and computational power); traditional methods fail to learn and understand the evolving behavior of the network to enable real-time and context-aware service discovery in such diverse and large-scale topologies of IoT networks. The Social IoT (SIoT) concept, which defines the relationships among the connected objects, can be exploited to extract established relationships between devices and enable trustworthy and context-aware services. In fact, SIoT expresses the possible connections that devices can establish in the network and reflect compatibility, trustworthiness, and so on. In this paper, we investigate the service discovery process in SIoT networks by proposing a low-complexity context-aware Graph Neural Network (GNN) approach to enable rapid and dynamic service discovery. Unlike the conventional graph-based techniques, the proposed approach simultaneously embeds the devices’ features and their SIoT relations. Our simulations on a real-world IoT dataset show that the proposed GNN-based approach can provide more concise clusters compared to traditional techniques, namely the Louvain and Leiden algorithms. This allows a better IoT network learning and understanding and also, speeds up the service lookup search space. Finally, we discuss implementing the GNN-assisted context-service discovery processes in novel smart city IoT-enabled applications.

In the last decade, community detection in dynamic networks has received increasing attention, because it can not only uncover the community structure of the network at any time but also reveal the regularity of dynamic networks evolution. Although methods based on the framework of evolutionary clustering are promising for dynamic community detection, there is still room for further improvement in the snapshot quality and the temporal cost. In this study, a dynamic community detection algorithm based on optional pathway guide pity beetle algorithm (DYN-OPGPBA), which is a novel dynamic community detection method based on the framework of evolutionary clustering, is proposed. We propose an improved PBA for community detection of the network at the first time step, including a discrete search strategy based on adjacent nodes, a closeness-based community modification strategy and a crowded community split strategy. Compared with many representative static community detection methods, the proposed method has some superior detection accuracy. A neighbour vector competition-based individual update strategy and an external population size restriction mechanism are also proposed for community detection at subsequent time steps. Results show that DYN-OPGPBA has a better balance between snapshot quality and temporal cost than two representative dynamic community detection methods.

Community structure has a more intuitive, physical meaning than how it has traditionally been perceived; this representation plays an important role in networks and is accordingly becoming widely adapted. Consequently, community detection has attracted increasing attention. Although scholars have proposed many community detection methods from different perspectives, due to the complexity, diversity and dynamic characteristics of networks, efficient community detection in many real networks remains a challenge. Inspired by fish school effect in real life, this paper envisions networks as an ecosystem and proposes a novel dynamic model that aims to reveal the communities in a more intuitive way. Relying on the new model, we design a community detection algorithm, known as community detection based on fish school effect (CDFSE). CDFSE has plentiful desirable properties: high-quality community detection, parameters free and notable scalability. To evaluate the performance of CDFSE, this paper employs two widely utilized evaluation metrics and twelve representative algorithms to test the effectiveness of the algorithm in both synthetic and real-world networks. The experimental results show that in most cases, CDFSE is superior to the comparison methods in terms of the quality of community detection.

Overlapping community structures are universal in complex networks. Multi-label propagation approach, as a method for detecting overlapping community, has been received widespread attention for the near linear time complexity. In recent years, some algorithms based on multi-label propagation approach have been proposed, but there is still room to improve the accuracy of these algorithms to detect overlapping communities. In this paper, a double-layer filtering strategy is proposed to help nodes filter labels, and the node centrality of each node according to the number of its neighbors and the number of edges between its neighbors is used to fix the node update sequence. Then based on the double-layer filtering strategy, the new fixed node update sequence and the previous multi-label propagation algorithm, a new algorithm called MLPA-DF is proposed. Theory analysis shows that MLPA-DF still has the near linear time complexity. And experimental results on the synthetic and real-world networks show that MLPA-DF has better performance than previous multi-label propagation algorithms (COPRA, SLPA and LPANNI) in terms of the accuracy of detecting overlapping communities without reducing stability.

Community detection can not only help people understand organizational structure and function of complex networks, but also attributes to many potential applications including targeted advertising and customer relationship management. Due to the low time complexity, the label propagation algorithm is widely used, but there is still room to improve the community quality and the detection stability. Inspired by resource allocation and local path similarity, we first give a new two-level neighbourhood similarity measure called TNS, and on this basis we propose an improved label propagation algorithm for community detection. In this new algorithm, the minimum distance and local centrality index are considered to select the initial community centers, to ensure that they are both important and far away from each other. In the process of forming initial community, we employ the new similarity measure and an optimization strategy of asynchronously updating labels according to node importance. To further improve the accuracy of community division, we introduce the label influence based on the new similarity measure to further optimize the community division of networks. The experimental results on both the artificial network and ten real-world networks show that our proposed algorithm has better comprehensive performance than several existing algorithms in terms of modularity, normalized mutual information and adjusted rand index.

Complex networks (CNs) have gained much attention in recent years due to their importance and popularity. The rapid growth in the size of CNs leads to more difficulties in the analysis of CNs tasks. Community Detection (CD) is an important multidisciplinary research area where many machine/deep learning-based methods have been applied to map CNs into a low-dimensional representation for extracting information similarity among members of CNs. Currently, Deep Learning (DL) is one of the promising methods to extract knowledge and learn information from high dimensional space and represent it in low dimensional space. However, designing an accurate and efficient DL-based CD method especially when dealing with large CNs is always an on-going research endeavor to pursue. Meta-Heuristic (MH) algorithms have shown their potentials in improving DL models in terms of solution quality and computational cost. In addition, parallel computing is a feasible solution for building efficient DL models. The algorithmic principle of MH is parallel in nature; however, its computation framework in DL training that is reported in the literature is not really implemented in a parallel computing setup. In this paper, we present a systematic review of CD in CNs from conventional machine learning to DL methods and point out the gap of applying DL-based CD methods in large CNs. In addition, the relevant studies on DL with parallel and MH approaches are reviewed and their implications on DL models are highlighted to prospect effective solutions to overcome the challenges of DL-based CD methods. We also point out research challenges in the field of CD and suggest possible future research directions.

Community detection in network-type data provides a powerful tool in analyzing and understanding real-world systems. In fact, community detection approaches aim to reduce the network’s dimensionality and partition it into a set of disjoint clusters or communities. However, real networks are usually corrupted with noise or outliers which affect the detected community structure quality. In this paper, a new robust community detection algorithm that is capable of recovering a clean or a smoothed version of the corrupted graph and detecting the correct community structure is introduced. The proposed approach combines robust principal component analysis (RPCA) and symmetric nonnegative matrix factorization (SymNMF) in a single optimization problem. The proposed problem is solved under the framework of alternating direction methods of multipliers (ADMM). In particular, the corrupted adjacency matrix is decomposed into a low-rank and sparse components using RPCA and the community structure is detected by applying SymNMF to the extracted low-rank component. Extensive experiments that have been conducted on real and simulated binary and weighted networks show that the proposed approach significantly outperforms existing algorithms in detecting the correct community structure even in grossly corrupted networks.

Experienced process plant personnel observe that corrective maintenance work on one asset may often be followed by corrective work on the same asset or connected assets within a short amount of time. This problem is referred to as a cascading failure. Confirming if these events are chronic is difficult given the number of assets and the volume of maintenance and operation data. If cascading events can be identified, preventative measures can be implemented to prevent those cascades, eliminating unnecessary corrective work. This project uses complex network analysis to identify cascading events and where co-occurrence of work is most frequent, in a process plant. Data is drawn from over 50,000 work orders for 5,655 pumps in a mining company over a five-year period. A complex network is produced by connecting assets based on the frequency of co-occurrence of work. Beside the advantages of the visualisation of complex networks, the method produces quantified measures, normalised degree, eigenvector centrality and betweenness centrality, which are used to identify assets with significant impact on other assets. Affected pumps are apparent as communities in the network. This analysis identifies pumps that are “super-spreaders”: pumps who experience corrective maintenance events which lead to corrective maintenance events on other pumps. The model can be tuned to different time windows, for example events within one or seven days. From these insights, changes can be made to operational, maintenance and recording practices to prevent re-occurrence. Of particular note in this data was the occurrence of self–loops in certain pumps and the prevalence of hidden failures in standby pumps.

With the fast development of network technology, great amount of data have been accumulated. Plenty of them are organized by using heterogeneous information networks (HIN). So mining heterogeneous information networks efficiently is very important. Node clustering is an essential part of this task. And several clustering algorithms have been proposed. As all these algorithms contain complicated optimization procedure and matrix calculation procedure, complexity of these algorithms is very high. To overcome the shortage described above, in this paper, a new clustering algorithm is proposed. In this algorithm, several parameters should be inputted. These parameters include a heterogeneous information network, meta-paths that are used and the names of target types. During the clustering procedure, a homogeneous network will be built by the proposed algorithm firstly. All the target objects of HIN are treated as nodes of this network. The instances of meta-paths are edges. After the homogeneous network is constructed, label propagation procedure can be performed. Then the clustering result will be obtained. Obviously, by using the proposed algorithm to perform clustering, the complex optimization procedure and matrix calculation procedure are eliminated. As the convergence rate of label propagation procedure is fast, the proposed algorithm is very efficient. Besides, we can find that label propagation procedure can be executed in parallel. Thus, the proposed algorithm is easy to be parallelized. In this situation, it is fit for processing large scale HIN based on server cluster. From experimental results, we can find that the proposed algorithm running faster than all the other algorithms for comparison.

Detecting community structure is an important problem in complex networks. Recently, the community detection method based on centers and neighbors (DCN) has been proposed, which is divided into two stages: community center point detection and label propagation. It has a better result on community detection of simple undirected graph than other algorithms. However, when there are two community centers connected directly, the method of DCN fails to find both centers. Based on the DCN algorithm, this paper proposes an optimization method based on visually-aided user interactions. By showing the local structure of the discovered center point through the force-directed layout, the potential community centers that are missing in DCN can be detected. We further propose the hierarchical visual clustering to assist users to detect more community centers easier. In addition, to make the propagation of labels more stable, we propose the multi-label propagation strategy based on importance which also preserves the labels proportion during propagation. The experimental results on both artificial and real-world networks demonstrate that our improved algorithm RefineDCN obtains better community detection results than the DCN algorithm.

In the context of cluster analysis and graph partitioning, many external evaluation measures have been proposed in the literature to compare two partitions of the same set. This makes the task of selecting the most appropriate measure for a given situation a challenge for the end user. However, this issue is overlooked in the literature. Researchers tend to follow tradition and use the standard measures of their field, although they often became standard only because previous researchers started consistently using them. In this work, we propose a new empirical evaluation framework to solve this issue, and help the end user selecting an appropriate measure for their application. For a collection of candidate measures, it first consists in describing their behavior by computing them for a generated dataset of partitions, obtained by applying a set of predefined parametric partition transformations. Second, our framework performs a regression analysis to characterize the measures in terms of how they are affected by these parameters and transformations. This allows both describing and comparing the measures. Our approach is not tied to any specific measure or application, so it can be applied to any situation. We illustrate its relevance by applying it to a selection of standard measures, and show how it can be put in practice through two concrete use cases.

Community detection is a crucial challenge in social network analysis. This task is important because it gives leads to study emerging phenomena. Indeed, it makes it possible to identify the different communities representing individuals with common interests and/or strong connections between them. In addition, it allows tracking the transformation of these communities over time. In this work, we propose a dynamic community detection approach called Attributes, Structure, and Messages distribution-based approach (ASMsg). In addition to the node attributes and the topological structure of the network, we use the rate of transferred messages as the key concept of this approach. Therefore, we obtain communities with similar members that are strongly connected and also frequently interacting. Furthermore, the proposed approach is able to detect all possible communities’ transformations even if the communities are overlapped. To demonstrate its efficiency, we widely test ASMsg on artificial and real-world dynamic networks and compare it with representative methods. The results show the superiority of our approach in terms of detected communities.

The community structure plays an indispensable role in developing the deep structure of complex networks. In recent years, some researchers have realized the importance of leader nodes in the community detection process. However, most of the existing leader-based algorithms only use the topological information of networks or attribute information to supplement the topological information, resulting in a significant loss of information integrity. In this paper, we propose a leader-based method that combines topological and attribute information (TALB), uses attribute information among nodes in the network to construct an attribute similarity matrix, and then combines it with network topological information to establish dependency relationships among nodes. As a result, a dependency tree is formed, and the final result of the community division is obtained. Experiments on synthetic networks and real networks show that our proposed method is more effective and practical than the existing leader-based algorithms.

Detecting communities in complex networks has been one of the most popular research areas in recent years. There have been many community detection algorithms proposed to date. However, the local information (cliques) of communities and the search efficiency of algorithm have not been considered both in previous studies. In this paper, we propose a novel local expansion algorithm for detecting overlapping communities based on cliques. The algorithm draws on the assumption that cliques are the core of communities, as the clique takes into account the local characteristics of the community. The proposed algorithm adopts a single node with the maximum density as an initial community to prevent the formation of a large number of near-duplicate community structures, which improves the search efficiency of the algorithm. In many experiments using computer-generated and real-world networks, the proposed algorithm based on this idea verifies that the algorithm is able to detect overlapping communities effectively. The experiment yields better community uncover results, and the time efficiency and the complexity of algorithm are also satisfactory.

A new type of residue interaction network named residue interaction energy network (RINN) is built. Then, a multi-objective optimization dynamic network community discovery algorithm T-DYNMOGA-Qw has been proposed to detect communities from dynamic RINN. T-DYNMOGA-Qw sets a threshold during the initialization process and optimizes weighted modularity Qw as the objective function. Setting the threshold can better find the stable structure in the dynamic RINN. The resolution limit of modularization has been broken by using objective function Qw. After Community detection from dynamic RINN of wild type of lipase (WTL) and its mutant 6B from 300K to 400K, it is found that the communities in 6B network can still maintain a tight structure even at higher temperature. Stable community is benefit to the heat resistance of lipase 6B. The hydrogen bonds between mutated Ser15 and Ser17, and the Glu20 with other residues improved the structure stability. The mutated L114P, M134E, M137P, and S163P enhance the rigidity of the flexible region and tighten the secondary structure, which stabilize the protein structure.

Community discovery (i.e. community detection) in signed networks is a division of nodes, such that the edges in the communities are positive and the edges between the communities are negative. Davis and Harary have solved the problem of community detection when a signed graph is balanced or weakly balanced. When the signed network is unbalanced, community detection becomes very complex. In this paper, we propose a novel memetic algorithm (MA) called MACD-SN for community partition (i.e. community detection) in signed networks. Firstly, we present a novel initialization algorithm used in initialization of MACD-SN. This method can accelerate the convergence rate of MACD-SN algorithm. Next, in addition to using frequently-used variation operation (in this paper, variation and mutation are interchangeable), this paper presents a novel crossover operation and a novel variation operation, which contributes to increasing the correctness of the MACD-SN algorithm's operation result and reduces its running time. Lastly, this paper proposes a new local search algorithm, which may enable the algorithm's result to jump away the local best result with a certain probability and draw near the global best result quickly. For testing the performance of MACD-SN algorithm, we have done many experiments using five kinds of synthetic signed networks and five real-world signed networks. The test outcomes show that the proposed algorithm is valid and efficient for signed network cluster partition (i.e. community detection).

The explosive growth of Internet of Things (i.e., IoT) terminal equipment makes its topology more complex, which leads to the increasing cost of network research. Recently, the implicit community structure is widely used to improve the efficiency of research. However, most of the non-overlapping community detection algorithms have some weakness, such as the large number of community detected and the obvious scale gap between communities. To address these abovementioned problems, we design a novel non-overlapping community detection algorithm, named as Pairing, Splitting and Aggregating algorithm (i.e., PSA). Firstly, in order to improve the accuracy of community division, a new node similarity index is designed to transform the network into a large number of similar node pairs. Secondly, based on the connected branches composed of similar node pairs, the network is further divided into several similar node sets. Thirdly, to balance the scale gap of different communities, the Grasshopper Optimization Algorithm, (i.e., GOA) is introduced to combine the local attribute (i.e., conductance) and global attribute (i.e., modularity) together to aggregate similar node sets into potential (or final) communities. Finally, the experimental results show that PSA not only controls the difference among communities well, but also outperforms the other four popular algorithms in terms of two metrics. Moreover, we propose a community-based resource discovery method (or scheme), named as Community-assisted Short-distance-query Resource Discovery algorithm (i.e., CSRD) to further verify the efficiency of PSA. The results show that the resource discovery efficiency of CSRD using PSA is better compared with other algorithms.

Most current survey papers classify community detection methods into broad categories and do not draw clear boundaries between the specific techniques employed by these methods. We survey in this paper all fine-grained community detection categories, the clustering methods that fall under these categories, and the techniques employed by these methods for optimizing each objective function. We provide methodology-based taxonomies that classify static and dynamic community detection methods into hierarchically nested, fine-grained, and specific classes. We classify the methods into the objective function they optimize. Each objective function class is classified into clustering categories. Each category is further classified into clustering methods. Methods are further classified into sub-methods and so on. Thus, the lowest subclass in a hierarchy is a fine-grained and specific method. For each method, we survey the different techniques in literature employed by the method. We empirically and experimentally compare and rank the different methods that fall under each clustering category. We also empirically and experimentally compare and rank the different categories that optimize a same objective function. In summary, the block-based, top-down divisive-based, random walk-based, and matrix eigenvector-based methods achieved good results. Finally, we provide fitness metrics for each objective function.

Graph embedding is an effective yet efficient way to convert graph data into a low dimensional space. In recent years, deep learning has applied on graph embedding and shown outstanding performance. Adjacency matrix is often taken as the storage data structure of graph. However, there are the problems of insufficient spatial proximity information in adjacency matrix. Therefore, this study proposes a deep community detection method which includes (1) matrix reconstruction method, (2) spatial feature extraction method and (3) community detection method. The original adjacency matrix in social network is reconstructed based on the opinion leader and nearer neighbors for obtaining spatial proximity matrix. The spatial proximity matrix can obtain subspace of the graph which can help convolution neural network easily and quickly extract the spatial localization. The spatial eigenvector of reconstructed adjacency matrix can be extracted by an auto-encoder based on convolution neural network for the improvement of modularity. In experiments, four open datasets of practical social networks were selected to evaluate the proposed method, and the experimental results show that the proposed deep community detection method obtained higher modularity than other deep learning methods. Therefore, the proposed deep community detection method can effectively detect high quality communities in social networks.

Detecting community structures is an important research topic in social network analysis. Unfortunately, the fundamental factors that drive the generation of social networks (i.e., the network topology and content) and community structures have not been well investigated. In this paper, according to the natural characteristics of social networks, we reveal that individual topics play a core role in community generation. If two individuals are in the same community and are interested in similar topics, it is more likely that a link will form between them. Otherwise, the probability of generating a link depends on the relationships between their communities and the topics they talk about. Based on the above observations, a novel generative community detection model is proposed that simulates the generation of the network topology and network content by considering individual topics. Moreover, our model utilizes a topic model to generate network content. The model is evaluated on two real-world datasets. The experimental results show that the community detection results outperform all the state-of-the-art baselines. In addition to accurate community detection results, we identify each individual topic distribution and the most popular users corresponding to different topics in each community.

Community detection is a critical issue in the field of complex networks. Recently, the nonnegative matrix factorization (NMF) method has successfully uncovered the community structure in the complex networks. However, this method has a significant drawback; most of community detection methods using NMF require the number of communities to be preassigned or determined the number of communities by searching for the best community structure among all candidates. To address this problem, in this paper, we use density peak clustering (DPC) to obtain the number of centers as the pre-defined parameter for nonnegative matrix factorization. However, due to sparse and high dimensional characteristics of complex networks, DPC cannot be used to detect community directly. To overcome this issue, we employ degree and hop of nodes as the density and distance indexes, respectively; we use NMF and Symmetric NMF to deal with linearly separable data and non-linearly separable data, respectively. Experimental results show that the proposed methods exhibit excellent performance on artificial and real-world networks and superior to the state-of-the-art methods which are the most common method for community detection of complex networks.

Community detection is a fundamental task in graph mining. Despite the fact that most of existing community detection methods are devoted to finding disjoint community structure, communities often overlap with each other and are recursively organized in a hierarchical structure in many real-world networks. Also, finding hierarchical overlapping community structure has significant implications in many real-world applications. Some of the few existing attempts suffer from the problem that the obtained community structure is sensitive to network changes as they are based heavily on one-hop node proximity to detect communities. To tackle this problem, we propose a robust hierarchical overlapping community detection method with Personalized PageRank (PPR), which is often regarded as a prevalent metric to measure node proximity globally. Specifically, motivated by the agglomerative hierarchical clustering method, we present an effective and efficient mechanism to merge small communities and form hierarchically organized overlapping communities. Experimental results on both synthetic and real-world networks corroborate the effectiveness and robustness of the proposed framework. In addition, we introduce how to make use of the detected community structure to perform various node proximity queries such as the top-k structural hole spanner query and the top-k heterogeneous node query, which can help us gain more insights on the underlying network.

The community structure detection in static networks often ignores the dynamic nature of the network and it is difficult to identify the evolution of community structure in dynamic networks. The community structure will converge or split as the nodes and edges change. Understanding the evolution of communities over time is an important issue in the study of social networks. Based on the characteristics of dynamic networks, this paper analyzed the influence of variables in dynamic networks structure. We proposed an Incremental algorithm with Coherent Neighborhood Propinquity in dynamic networks. The algorithm considered the direct and indirect effects of changing nodes in their previous communities. We also considered the coherent neighborhood propinquity and improved the influence range of variable nodes. Comparing with the traditional algorithms, the experimental results showed that the proposed algorithm has better performance and less running time.

Identification of community structures is essential for characterizing and analyzing complex networks. Having focusing primarily on network topological structures, most existing methods for community detection ignore two types of non-topological relationships among nodes, i.e., pairwise “must-link” constraints among pairs of nodes and labels of nodes, such as functions they may have. Here, we present a novel semi-supervised and active learning method for community detection to integrate these two types of information of a network so as to increase the accuracy of community identification. Our new method will honor the “must-link” relationship without introducing new parameters and is efficient with a guaranteed convergence. An essential component of the method is a linear representation that is particularly suited to an active learning to help select the most critical nodes that impact community discovery. We present results from extensive experiments on synthetic and real networks to show the superior performance of the new methods over the existing approaches.

Human activities are usually collective, so clustering has become an important feature of human behavior. This paper studied the evolution of the community in the process of public opinion propagation so as to put forward a public opinion evolution model for the network community number. This study proposed the community number evolution model of public opinion based on stochastic competitive learning, and the proposed model consists of an increase in the number of communities and a decrease in the number of communities. The highlight of this model is that on the one hand, it realizes the research on the evolution of public opinions on the dynamic network; on the other hand, unlike other public opinion evolution models, this model pays attention to the community number increase and decrease rules in the evolution of public opinions. Then, as an extension of the community number evolution model of public opinion, the community number prediction model had been proposed. Based on Twitter data from the 2017 London Bridge attack, the proposed models were validated by experiments. In the verification section of this paper, two methods had been introduced as a comparison. The experimental results show that the community number evolution model of public opinion is correct.

Community is an important topological characteristic of complex networks, which is significant for understanding the structural feature and organizational function of networks, and community detection has recently attracted considerable research effort. Among community detection methods, label propagation technology is widely used because of its linear time complexity. However, due to the randomness of the node order of label updating and the order of label launching in label propagation, the instability of community detection approaches based on label propagation becomes a challenge. In this paper, a new label propagation algorithm, Node Ability based Label Propagation Algorithm (NALPA), is proposed to discover communities in networks. Inspired from human society and radar transmission, we design four node abilities (propagation ability, attraction ability, launch ability and acceptance ability), label influence and a new label propagation mechanism to deal with the instability and enhance the efficiency. Experimental results on 42 synthetic and 14 real-world networks demonstrate that NALPA outperforms state-of-the-art approaches in most cases. In a case study, NALPA is applied to a drug network in Traditional Chinese Medicine (TCM) and can discover the drug communities where drugs have similar efficacy for treating Chronic GlomeruloNephritis (CGN).