Hierarchical community detection via rank-2 symmetric nonnegative matrix factorization.

Q1 Mathematics

Computational Social Networks Pub Date : 2017-01-01 Epub Date: 2017-09-08 DOI:10.1186/s40649-017-0043-5

Rundong Du, Da Kuang, Barry Drake, Haesun Park

{"title":"Hierarchical community detection via rank-2 symmetric nonnegative matrix factorization.","authors":"Rundong Du, Da Kuang, Barry Drake, Haesun Park","doi":"10.1186/s40649-017-0043-5","DOIUrl":null,"url":null,"abstract":"Background: Community discovery is an important task for revealing structures in large networks. The massive size of contemporary social networks poses a tremendous challenge to the scalability of traditional graph clustering algorithms and the evaluation of discovered communities.Methods: We propose a divide-and-conquer strategy to discover hierarchical community structure, nonoverlapping within each level. Our algorithm is based on the highly efficient rank-2 symmetric nonnegative matrix factorization. We solve several implementation challenges to boost its efficiency on modern computer architectures, specifically for very sparse adjacency matrices that represent a wide range of social networks.Conclusions: Empirical results have shown that our algorithm has competitive overall efficiency and leading performance in minimizing the average normalized cut, and that the nonoverlapping communities found by our algorithm recover the ground-truth communities better than state-of-the-art algorithms for overlapping community detection. In addition, we present a new dataset of the DBLP computer science bibliography network with richer meta-data and verifiable ground-truth knowledge, which can foster future research in community finding and interpretation of communities in large networks.","PeriodicalId":52145,"journal":{"name":"Computational Social Networks","volume":"4 1","pages":"7"},"PeriodicalIF":0.0000,"publicationDate":"2017-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1186/s40649-017-0043-5","citationCount":"10","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computational Social Networks","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1186/s40649-017-0043-5","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2017/9/8 0:00:00","PubModel":"Epub","JCR":"Q1","JCRName":"Mathematics","Score":null,"Total":0}

引用次数: 10

Abstract

Background: Community discovery is an important task for revealing structures in large networks. The massive size of contemporary social networks poses a tremendous challenge to the scalability of traditional graph clustering algorithms and the evaluation of discovered communities.

Methods: We propose a divide-and-conquer strategy to discover hierarchical community structure, nonoverlapping within each level. Our algorithm is based on the highly efficient rank-2 symmetric nonnegative matrix factorization. We solve several implementation challenges to boost its efficiency on modern computer architectures, specifically for very sparse adjacency matrices that represent a wide range of social networks.

Conclusions: Empirical results have shown that our algorithm has competitive overall efficiency and leading performance in minimizing the average normalized cut, and that the nonoverlapping communities found by our algorithm recover the ground-truth communities better than state-of-the-art algorithms for overlapping community detection. In addition, we present a new dataset of the DBLP computer science bibliography network with richer meta-data and verifiable ground-truth knowledge, which can foster future research in community finding and interpretation of communities in large networks.

Abstract Image

查看原文本刊更多论文

基于秩-2对称非负矩阵分解的分层社团检测。

背景:社区发现是揭示大型网络结构的重要任务。当代社交网络的巨大规模对传统的图聚类算法的可扩展性和发现社区的评估提出了巨大的挑战。方法:提出了一种分而治之的策略来发现分层的社区结构，每个层次内部不重叠。该算法基于高效的秩-2对称非负矩阵分解。我们解决了几个实现挑战，以提高其在现代计算机体系结构上的效率，特别是对于代表广泛社会网络的非常稀疏的邻接矩阵。结论:实验结果表明，我们的算法在最小化平均归一化切割方面具有竞争力的整体效率和领先的性能，并且我们的算法发现的非重叠社区比最先进的重叠社区检测算法更好地恢复了真实社区。此外，我们提出了一个新的DBLP计算机科学书目网络数据集，它具有更丰富的元数据和可验证的基础事实知识，可以促进未来在大型网络中社区发现和社区解释的研究。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Computational Social Networks Mathematics-Modeling and Simulation

自引率

0.00%

发文量

审稿时长

13 weeks

期刊介绍： Computational Social Networks showcases refereed papers dealing with all mathematical, computational and applied aspects of social computing. The objective of this journal is to advance and promote the theoretical foundation, mathematical aspects, and applications of social computing. Submissions are welcome which focus on common principles, algorithms and tools that govern network structures/topologies, network functionalities, security and privacy, network behaviors, information diffusions and influence, social recommendation systems which are applicable to all types of social networks and social media. Topics include (but are not limited to) the following: -Social network design and architecture -Mathematical modeling and analysis -Real-world complex networks -Information retrieval in social contexts, political analysts -Network structure analysis -Network dynamics optimization -Complex network robustness and vulnerability -Information diffusion models and analysis -Security and privacy -Searching in complex networks -Efficient algorithms -Network behaviors -Trust and reputation -Social Influence -Social Recommendation -Social media analysis -Big data analysis on online social networks This journal publishes rigorously refereed papers dealing with all mathematical, computational and applied aspects of social computing. The journal also includes reviews of appropriate books as special issues on hot topics.