Overlapping community detection in heterogeneous social networks via the user model

doi:10.1016/j.ins.2017.11.055

Information Sciences

Volume 432, March 2018, Pages 164-184

https://doi.org/10.1016/j.ins.2017.11.055 Get rights and content

Abstract

Clustering users with more common interests who interact frequently on social networking sites has attracted much attention from researchers due to the high economic value and further application prospects. Community detection is a widely accepted means of dealing with the challenge of clustering users, but conventional methods are inadequate since there are billions of vertices and various relations in social media. Through the user model, a heterogeneous network containing both undirected and directed edges is built in this study to exactly simulate a social network. A novel approach for overlapping community detection in a heterogeneous social network (OCD-HSN) is proposed, which contains seed selecting and community initializing and expanding to accurately and efficiently unfold modules in parallel. Experimental results on artificial and real-world social networks demonstrate the higher accuracy and lower time consumption of the proposed scheme compared with other existing state-of-the-art algorithms.

Introduction

With the rapid development of Web 2.0 technology, the widespread acceptance and utilization of social networking sites (SNSs) have changed the modes of communication by facilitating the ability of individuals from diverse areas to interact and share interests. Social networks such as Twitter, Facebook, Google+, Orkut, LinkedIn, Friendster and Multiply have attracted billions of users, and have also received attention from many researchers and media. Whereas these sites offer potential benefits to users [22], they also pose a challenging problem of information overload and place higher requirements on community detection.

Communities, also known as clusters or modules, are the sets of vertices with common properties or the same roles in the network [12]. Community detection is a process of identifying all communities in a network. A large variety of techniques is available to carry out this single task. For more information, see the recent survey by Schaffer [33]. The ability to unfold these sub-structures in a social network can provide insight into how the network topology and functions affect each other, and can further provide theoretical support for interest prediction, friend recommendation, and event evolution.

Entities in social systems are treated as vertices, and edges are connected between vertices based on various relations. With such a representation, the theories of complex networks can be applied to solve many specific problems; this has currently become the leading method in pertinent research efforts [13]. How to construct appropriate complex networks based on the characteristics of social media and explore effective community detection algorithms in large-scale social networks, however, has become challenging. This study aims at meeting this challenge.

The user model, also called the user profile, as a mirror of a real user in the cyber-world, is constructed to depict basic user attributes, interests, and social relations. Although the Vector Space Model [19] has been widely adopted to model users, it suffers from several critical weaknesses. It models users in a bloated form, has no uniform representation standard, and cannot implement semantic expansion. To address these shortcomings, the semantic ontology representation [9] is selected in this study to model users’ semantic interests, and social relations among users are expressed independently of users’ interests.

The Pearson correlation coefficient is applied to calculate the interest similarity value between user models on the corresponding layer of the semantic ontology, and the interest similarity values of multiple layers are mixed using the hyperbolic tangent function. We combine interest similarities and social interactions in the same network to construct a heterogeneous network that simultaneously contains undirected and directed edges.

In this study, it is demonstrated that the community detection in heterogeneous networks using seed set expansion can not only help mine overlapping and hierarchical communities, but can also calculate membership degrees of overlapping vertices in different communities. To the best of our knowledge, this is the first attempt to study overlapping community detection in heterogeneous social networks (OCD-HSN) containing both undirected and directed edges.

The main characteristics and innovations of this study are the following:

•
The interest similarities and social interactions among user models are both included in the same heterogeneous network.
•
A novel approach is proposed to detect communities in heterogeneous networks containing both undirected and directed edges.
•
Multiple seeds are selected in the heterogeneous network to facilitate community detection in parallel.
•
Each community is initialized to a vertex set according to the seed vertex characteristic to improve community accuracy and accelerate detection speed.
•
The election principle is applied to add and remove multiple vertices during iterative expanding of communities, to further accelerate community detection.

To assess the accuracy and effectiveness of our method, we have compared it with two other existing state-of-the-art algorithms. Experimental results on both artificial and real-world social networks indicate that our proposed approach outperforms the other two methods with higher accuracy and lower time complexity. With these advantages, our approach can well satisfy the application requirements of large-scale social networks with billions of vertices.

The rest of this paper is organized as follows. In Section 2, we summarize related work on the techniques of user modeling and community detection. We introduce the user modeling method and build a heterogeneous social network in Section 3. In Section 4, the proposed community detection approach is described in depth. The experimental results of the proposed scheme are compared and analyzed in Section 5. Finally, we conclude the paper in Section 6 by discussing plans for future work.

Section snippets

Related work

Different from traditional complex networks, social networks express users of SNSs and the relations among them. By analyzing such networks, not only is it possible to gain insights into social phenomena and processes that take place in the real world, but one can also extract actionable knowledge that can be beneficial in retrieval tasks and information management, such as online content navigation and recommendation systems.

Heterogeneous social networks

In this section, we generate a heterogeneous social network to provide the foundation for community detection. First, each user is exhibited as a user model with semantic interests and social interactions by analyzing the features of users. We then calculate the interest similarities and connect user models based on the social relations among them, yielding a graph with vertices representing user models and edges representing various relations.

Community detection approach

In this section, we first present the framework of our proposed approach. Next, we describe in detail the multiple seeds selection strategy, overlapping community detection algorithm, and membership degree calculation method in the heterogeneous social network. Finally, we reveal how to discover the hierarchical module structure and analyze the time complexity.

Experiments

In this section, we first introduce four assessment criteria that are extended from traditional unweighted and undirected networks, and two state-of-the-art methods that are employed for comparison with our proposed approach. Second, we present the experimental results on both synthetic and real-world heterogeneous social networks. Finally, we discuss and analyze the experimental results.

Conclusions and Future Work

In this study, we propose an overall framework for overlapping community detection in the heterogeneous social network. First, we construct a heterogeneous social network using the similarity calculation of multiple user interests and the consideration of social interaction relations, where the user’s interests and social features are depicted by a user model as a vertex in the network, to provide a new perspective for social computing. We then present an integrative approach for community

Acknowledgments

This work is partially funded by the National Key Research and Development Program of China (No. 2017YFC0907505), the National Natural Science Foundation of China (Nos. 61772128 and 61303096), the Fundamental Research Funds for the Central Universities (No. 16D111208), the Shanghai Natural Science Foundation (No. 17ZR1400200), and the Xinjiang Social Science Foundation (No. 2015BGL100).

References (37)

A. Arenas et al.
Synchronization processes in complex networks
Phys. D
(2006)
S. Calegari et al.
Personal ontologies: generation of user profiles based on the YAGO ontology
Inf. Process Manage
(2013)
S. Fortunato
Community detection in graphs
Phys. Rep.
(2010)
K. Ikeda et al.
Twitter user profiling based on text and community mining for market analysis
Knowl. Based Syst.
(2013)
S.E. Schaeffer
Graph clustering
Comput. Sci. Rev.
(2007)
H. Shen et al.
Detect overlapping and hierarchical community structure in networks
Physica A
(2009)
J. Zheng et al.
Neighborhood-user profiling based on perception relationship in the micro-blog scenario
Web Semant.
(2015)
Baidu encyclopedia,...
Tencent microblog open platform,...
V. Agarwal et al.
A collaborative filtering framework for friends recommendation in social networks based on interaction intensity and adaptive user similarity
Soc. Netw. Anal. Min.
(2013)

R. Andersen et al.

Local graph partitioning using pagerank vectors

Proceedings of the 47th Annual IEEE Symposium on Foundations of Computer Science (FOCS’06)

(2006)

A. Arenas et al.

Synchronization reveals topological scales in complex networks

Phys. Rev. Lett.

(2006)

S. Asur et al.

An event-based framework for characterizing the evolutionary behavior of interaction graphs

ACM Trans. Knowl. Discov. Data (TKDD)

(2009)

V.D. Blondel et al.

Fast unfolding of communities in large networks

J. Stat. Mech: Theory Exp.

(2008)

L.M. Collins et al.

Omega: a general formulation of the rand index of cluster recovery suitable for non-disjoint solutions

Multivariate Behav. Res.

(1988)

S.G. Esparza et al.

Mining the real-time web: a novel approach to product recommendation

Knowl. Based Syst.

(2012)

G. Ghoshal et al.

Social system as complex networks

Soc. Netw. Anal. Min.

(2014)

M. Girvan et al.

Community structure in social and biological networks

Proc. Natl. Acad. Sci.

(2002)

Cited by (42)

Interaction behavior enhanced community detection in online social networks
2024, Computer Communications
Community detection in online social networks (OSNs) has attracted a great deal of attention in recent years due to its potential prospects in applications such as recommendation systems, spam detection, targeted advertising, etc. However, community detection methods that have been proposed so far rely mostly on static information, such as the topology of the network, the attributes of the users, etc. In this paper, we propose a framework that enhances community detection with user interaction behaviors in OSNs since such behaviors should be important factors in characterizing the communities. In the framework, the various features of user interaction behaviors are expressed in the form of interaction indices through utilizing the cumulative distribution function and interaction divergence. The indices are then transformed into edge weights through a logistic regression model and combined with the static information to construct a weighted graph for the OSN. Finally, community detection is realized by applying a modularity maximization algorithm. A comprehensive evaluation using diverse real-world datasets shows that our proposed framework has superior performance over comparable ones in deriving high-quality communities. Critical metrics, such as modularity, average clustering coefficient, and average conductance, are evaluated to demonstrate the effectiveness of our approach. Furthermore, the synergy between network topology and user interaction behavior in community detection is also investigated.
Sparse representation for heterogeneous information networks
2023, Neurocomputing
A complex network is a fundamental tool to describe real-world complex systems, with most real-world systems containing multiple object types and relationships that can be described as heterogeneous information networks. However, with the increasing network complexity, understanding the complex patterns and finding the meta paths or meta-structures of the heterogeneous information networks has become challenging. This paper proposes a sparse representation for heterogeneous information networks and extracts the heterogeneous information atoms that describe the basic connection pattern of the original heterogeneous information network. The heterogeneous information atoms help extract the main meta-paths or meta-structures and understand the complex patterns of the original heterogeneous information network. Furthermore, the heterogeneous information networks can be decomposed, dimension-reduced, and reconstructed through the heterogeneous information atoms. Extensive experimental results demonstrate that heterogeneous information atoms and sparse coding represent the basic connection pattern of real-world heterogeneous information networks. Indeed, the developed method can reconstruct a network with a recovery exceeding 90%.
CEO: Identifying Overlapping Communities via Construction, Expansion and Optimization
2022, Information Sciences
Citation Excerpt :
However, increasing evidence shows that many real-world systems are characterized by statistics of overlapping communities where a node may belong to multiple communities [4]. Mainstream overlapping community detection algorithms can be classified into local expansion methods [5], clique percolation methods [6], link partitioning methods [7], agent-based dynamical methods [8], fuzzy detection methods [9], multi-objective evolutionary methods [10], learning-based methods [11], etc. In addition, by performing on multiple groups of query nodes, community search methods can generate communities with overlaps [12].
Overlapping communities are ubiquitous in real-world systems. For overlapping community detection, local expansion methods excel in scalability and efficiency yet have poor tolerance to low-quality seeds and communities. Based on our previous work, we introduce a more robust local-expansion-based overlapping community detection algorithm, named CEO, performing Construction, Expansion and Optimization sub-processes. To solve the poor fault tolerance problem, CEO discards low-quality seeds and communities in each sub-process based on optimizing node memberships. CEO was compared to thirteen noted algorithms by examining the performance on five groups of artificial networks and sixteen real-world networks with ground-truth communities. Experimental results showed CEO performs the best in identifying overlapping communities, which verifies the effectiveness of discarding low-quality seeds and communities in solving the poor fault tolerance problem.
Parallel and distributed paradigms for community detection in social networks: A methodological review
2022, Expert Systems with Applications
Community detection in social networks is the process of identifying the cohesive groups of similar nodes. Detection of these groups can be helpful in many applications, such as finding networks of protein interaction in biological networks, finding the users of similar mind for ads and suggestions, finding a shared research field in collaborative networks, analyzing public health, future link prediction in social networks, analyzing criminology, and many more. However, with the increase in the number of profiles and content shared on social media platforms, the analysis is often time-consuming and exhaustive. In order to speed up and optimize the community detection process, parallel processing and Shared/Distributed memory techniques are widely used. Despite community detection has widespread use in social networks, no attempt has ever been made to compile and systematically discuss research efforts on the emerging subject of identifying parallel and distributed methods for community detection in social networks. Most of the surveys described the serial algorithms used for community detection. Our survey work comes under the scope of new design techniques, exciting or novel applications, components or standards, and applications of an educational, transactional, and co-operational nature. This paper accommodates and presents a systematic literature review with state-of-the-art research on the application of parallel processing and Shared/Distributed techniques to determine communities for social network analysis. Advanced search strategy has been performed on several digital libraries for extracting several studies for the review. The systematic search landed in finding 3220 studies, among which 65 relevant studies are selected after conducting various screening phases for further review. The application of parallel computing, shared memory, and distributed memory on the existing community detection methodologies have been discussed thoroughly. More specifically, the central significance of this paper is that a systematic literature review is conducted to gather the relevant studies from different digital libraries and gray literature. Then, different parametric values of each selected study are appropriately compared. Moreover, the need for further research to speed up the process of community formation in parallel and shared approaches has been pinpointed more suitably. A pictorial glance of this paper is depicted as follows:
Social influence based community detection in event-based social networks
2020, Information Processing and Management
Citation Excerpt :
Community detection is an important problem in social network analysis and it has received a great deal of attention from many disciplines (Fortunato, 2010; He, Li, Soundarajan, & Hopcroft, 2018; Newman, 2004; Wu, Kwong, Zhou, Jia, & Gao, 2018). The ability to detect internally connected communities in social networks can provide supports for real-world applications, such as friend recommendation and interest prediction (Huang et al., 2018). In general, a community in social networks represents a group of people who are connected internally (Newman, 2004).
In this paper, we focus on the problem of discovering internally connected communities in event-based social networks (EBSNs) and propose a community detection method by utilizing social influences between users. Different from traditional social network, EBSNs contain different types of entities and links, and users in EBSNs have more complex behaviours. This leads to poor performance of the traditional social influence computation method in EBSNs. Therefore, to quantify the pairwise social influence accurately in EBSNs, we first propose to compute two types of social influences, i.e., structure-based social influence and behaviour-based social influence, by utilizing the online social network structure and offline social behaviours of users. In particular, based on the specific features of EBSNs, the similarities of user preference on three aspects (i.e., topics, regions and organizers) are utilized to measure the behaviour-based social influence. Then, we obtain the unified pairwise social influence by combining these two types of social influences through a weight function. Next, we present a social influence based community detection algorithm which is referred to as SICD. In SICD, inspired by the nonlinear feature learning ability of the autoencoder, we first devise a neighborhood based deep autoencoder algorithm to obtain nonlinear community-oriented latent representations of users, and then utilize the k-means algorithm for community detection. Experimental results conducted on real-world dataset show the effectiveness of our proposed algorithm.
Identification of topical subpopulations on social media
2020, Information Sciences
Citation Excerpt :
As the processing and storage costs of individual topical profiles is high, the application of such an approach in ad-hoc search setup is non-trivial. More recently, Huang et al. [20] similarly mapped the content associated with users on social media onto an ontology. They inferred topical similarity between users, and co-represented this information alongside social relations in a heterogeneous joint graph for the purpose of overlapping community detection.
We tackle a major challenge of information filtering on social media (SM): rather than address the general question “what are people talking about on SM?”, we consider a finer question, “what are ... talking about on SM?”, where ... stands for some subpopulation of SM users of interest. We take a set expansion approach, where a seed of example members of the target subpopulation is initially defined, and additional SM users who belong to that subpopulation are identified, thus enabling the effective tracking of relevant information that pertains to that subpopulation on SM. Specifically, the Personalized PageRank (PPR) random walk measure is iteratively applied to detect additional members of the subpopulation based on their structural similarity to the seed set within the social media graph. There are several main contributions of this work. We outline Splash PPR, an efficient distributed computation of PPR adapted for potentially large seed sets and very large SM graphs. Using Splash PPR, we examine and tune graph representations towards the retrieval of two subpopulations from Twitter, namely human rights Activists, and Machine Learning practitioners. We believe this work is first to introduce and evaluate a generic framework for subpopulation identification at scale.

View all citing articles on Scopus

View full text

Overlapping community detection in heterogeneous social networks via the user model

Abstract

Introduction

Section snippets

Related work

Heterogeneous social networks

Community detection approach

Experiments

Conclusions and Future Work

Acknowledgments

Phys. D

Inf. Process Manage

Phys. Rep.

Knowl. Based Syst.

Comput. Sci. Rev.

Physica A

Web Semant.

A collaborative filtering framework for friends recommendation in social networks based on interaction intensity and adaptive user similarity

Soc. Netw. Anal. Min.

Local graph partitioning using pagerank vectors

Proceedings of the 47th Annual IEEE Symposium on Foundations of Computer Science (FOCS’06)

Synchronization reveals topological scales in complex networks

Phys. Rev. Lett.

An event-based framework for characterizing the evolutionary behavior of interaction graphs

ACM Trans. Knowl. Discov. Data (TKDD)

Fast unfolding of communities in large networks

J. Stat. Mech: Theory Exp.

Omega: a general formulation of the rand index of cluster recovery suitable for non-disjoint solutions

Multivariate Behav. Res.

Mining the real-time web: a novel approach to product recommendation

Knowl. Based Syst.

Social system as complex networks

Soc. Netw. Anal. Min.

Community structure in social and biological networks

Proc. Natl. Acad. Sci.