基于网络层加权的多层复杂网络社区检测算法

Multi-layer Complex Network Community Detection Algorithm Based on Network Layer Weighting

社区检测被用于通过观察拓扑结构来寻找网络的最合理的分区,是多层复杂网络分析的一个重要任务。针对多层复杂网络的社区检测问题,该文提出了一种基于网络层加权的局部社区检测算法(MWLCD)。首先针对不同类型的网络,该算法使用超参数结合两种不同的加权方案量化了多层复杂网络中不同网络层的权值,然后引入了局部密度概念来确定种子节点附近的核心节点,以此避免了局部社区检测的社区划分质量依赖于种子节点所在位置优劣的问题。该算法基于社区核心节点将剩余节点依据其是否存在于社区内分为边界节点和外壳节点以及网络的未探索部分节点,并在外壳节点集中选取最适合该社区的候选节点放入社区同时不断更新三个节点集直到所有的社区划分完毕,然后采用多层模块度函数评估社区划分的质量。实验结果表明:在4个公开的多层复杂网络数据集上,MWLCD算法划分的社区可以取得更好的多层模块度值。

Community detection is used to find the most reasonable partition of the network by observing the topological structure,which is an important task of multi-layer complex network analysis.Aiming at the problem of community detection in multi-layer complex networks,we propose a local community detection algorithm(MWLCD)based on network layer weighting.The algorithm first quantifies the weights of different network layers in a multi-layer complex network using hyperparameters and two different weighting schemes for different types of networks,and then introduces the concept of local density to determine the core nodes near the seed nodes,which avoids the problem that the quality of community division of local community detection depends on the location of seed nodes.Based on the core node of the community,the algorithm divides the remaining nodes into boundary nodes,shell nodes,and unexplored part of the network according to whether they exist in the community,and selects the most suitable candidate nodes for the community from the shell nodes and puts them in the community while constantly updating three node sets until all the communities are divided,and then the multi-layer modularity function is used to evaluate the quality of the community division.The experiment shows that on the four public multi-layer complex network data sets,the communities divided by the MWLCD algorithm can obtain better multi-layer modularity values.