摘要
寻找网络中连接紧密的、稳定的社区,对网络大数据的挖掘和分析具有重要的意义和价值.节点属性和网络拓扑对社区发现都有重要的影响,由于真实网络中的节点属性维度大,找寻重要属性困难,而且和深层次的结构信息又不易进行高效整合以进行社区划分.为了有效地提取节点的重要属性信息,并和局部链接拓扑信息深入融合,根据矩阵分解,提出了基于特征选择和属性网络嵌入的社区发现算法.首先采用节点的联合相似度潜在表征指导特征选择,筛选出重要的属性后与原拓扑组成新网络,然后将新网络通过融合邻居信息的属性网络表征学习映射成节点低维向量,最后对该嵌入向量进行聚类从而实现社区划分.在真实网络数据集上与其他代表性算法进行比较,实验结果表明所提算法具有良好的特征选择性能和社团划分性能.
It is of great significance and value to find the closely connected and stable community in the network for the mining and analysis of network big data.Due to the large dimension of node attributes in the real world,it is difficult to find important node attributes;in addition,node attributes and network topology both have an important influence on community discovery,but it is difficult to integrate the two to divide the community.In order to integrate the important attribute information and local link topology information of nodes through matrix decomposition,a community discovery method based on feature selection and attribute network embedding is proposed.Firstly,feature selection is guided by the potential representation of joint similarity of nodes,and important attributes are screened out to form a new network with the original topology.Secondly,the new network is mapped into a low-dimensional vector through attribute network representation learning combined with neighbor information,and the embedding vector integrates topological structure and important attribute information.Finally,the embedding vector is clustered to realize community division.Compared with other representative algorithms on real network data sets,the experimental results show that this algorithm has good performance of feature selection and community detection.
作者
徐新黎
尹晶
肖云月
龙海霞
XU Xin-li;YIN Jing;XIAO Yun-yue;LONG Hai-xia(College of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310023,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2023年第5期939-946,共8页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61773348)资助
浙江省公益科技计划项目(LGG20F020017)资助。
关键词
特征选择
表征学习
属性网络嵌入
社区发现
矩阵分解
feature selection
representation learning
attributed network embedding
community detection
matrix decomposition