当前社区发现领域存在诸多静态社区划分算法,而其划分结果的不稳定性和较高的算法复杂度已经不能适应如今规模庞大,变化频繁的网络结构.为解决传统静态算法这一局限性,提出了一种利用模块度优化的增量学习算法,将网络结构的变化划分成边变化、点变化两种基本操作,在对“模块度最大化”的规则指导下实现网络结构的增量学习.实验表明,该算法在保证原有社区划分结果的前提下,可以将新变化的节点快速划分进已有社区,并使得模块度与静态算法重新计算模块度相近,节省了时间,保持了社区划分的实时性.%There are many static algorithms in the community detection fields but very few of them are able to fit into the current network circumstances where network sizes are getting larger and small-scale changes are getting more frequent.To deal with the limitation,this paper proposes an incremental learning algorithm of community detection based on modularity,which takes consideration of two basic kinds of community operations,edge changes and point changes,and the incremental learning process is carried out with guidance of the principle of modularity maximization.Experimental evaluation shows that the incremental learning algorithm is capable of partitioning new timely changed nodes into existed communities rapidly without any devastation to primitive divisions.Meanwhile,the result is proved to get close to that gained by static algorithm thus saving time and keeping real-time.
展开▼
