摘要
针对有向加权网络节点类型多、数据量大,导致重叠检测精准度不高的问题,提出一种基于深度迁移的检测算法。利用深度迁移学习规则求解上一迁移任务节点间聚合、分离度关系,获取下一任务节点重叠知识,根据二者比值求得中心度值,中心度最高的节点与其它节点存在强紧密连接关系。考虑到有向加权网络范围大、节点数量多,计算中心节点的网络社区隶属度值越高表明社区内重叠节点数量越多。赋予重叠社区内相邻节点和数据边不同权重,定义两点间邻域重叠比,重叠比越大、数据间连接程度越强,重叠概率越高。仿真结果证明,所提方法检出率高,负载率低,耗用时间短,应用性能强。
A deep migration-based detection algorithm is proposed to address the problem of low accuracy in overlap detection due to the large number of node types and data volume in directed weighted networks.Firstly,deep transfer learning rules were used to solve the polymerization and separation relationship between previous migration task nodes,and thus to obtain the overlapping knowledge of the next task node.According to the ratio of the two,the node with maximum centrality had a strong and tight connection with other nodes.Considering that the directed weighted network has a large range and a large number of nodes,the higher the network community membership value of the center node,the more overlapping nodes in the community.Different weights were given to adjacent nodes and data edges in overlapping community,and then the overlap ratio of the neighborhood between the two points.The higher the overlap ratio,the stronger the connectivity between data,and the higher the probability of overlap.Simulation results prove that the proposed method has high detectable rate,low load rate,low time consumption and strong application performance.
作者
王小红
刘琴
WANG Xiao-hong;LIU Qin(Computer College,Qinghai Nationalities University,Xining Qinghai 810000,China)
出处
《计算机仿真》
北大核心
2023年第9期492-496,共5页
Computer Simulation
关键词
深度迁移
聚合度
隶属度
相邻节点
邻域重叠比
Deep migration
Degree of polymerization
Degree of membership
Adjacent node
Neighborhood overlap ratio
