Fluid Communities: A Community Detection Algorithm
Fluid Communities: A Community Detection Algorithm
复制标题
流动社区：社区检测算法
DOI：
--
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
T. Suzumura
中科院分区：
文献类型：
--
作者：
Ferran Parés;D. García;Armand Vilalta;Jonatan Moreno;E. Ayguadé;Jesús Labarta;Ulises Cortés;T. Suzumura
社区检测算法是一系列无监督图挖掘算法，其将顶点分组为簇（即，社区）。这些算法提供了深入了解网络的结构和组成it. In本文中，我们提出了一种新的社区检测算法的基础上的简单的想法，流体在环境中相互作用，扩大和收缩接触。Fluid Communities算法基于传播方法，这是在计算成本和可扩展性方面最有效的社区检测方法。同时，它发现的社区的质量接近当前最先进的社区检测算法，并显着优于标签传播算法（LPA）的上级。虽然所有先前提出的基于传播的算法只能为给定的图产生单个聚类，但流体社区算法可以识别可变数量的社区。因此，该算法是一个独特的和可扩展的工具，用于分析大规模图的拓扑结构，在多个程度的粒度。关键词-社区检测，网络分析，图挖掘，无监督学习。
Community detection algorithms are a family of unsupervised graph mining algorithms which group vertices into clusters (i.e., communities). These algorithms provide insight into both the structure of a network and the entities that compose it. In this paper we propose a novel community detection algorithm based on the simple idea of fluids interacting in an environment, expanding and contracting in contact with one another. The Fluid Communities algorithm is based on the propagation methodology, the most efficient approach to community detection in terms of computational cost and scalability. At the same time, the quality of the communities it finds is close to that of the current state-of-the-art community detection algorithms, and significantly superior to the Label Propagation Algorithm (LPA). While all previously proposed propagation-based algorithms can only produce a single clustering for a given graph, the Fluid Communities algorithm can identify a variable number of communities. As a result, the proposed algorithm represents a distinct and scalable tool for analyzing the topology of large scale graphs at multiple degrees of granularity. Keywords—Community Detection, Network Analysis, Graph Mining, Unsupervised Learning.