从消息传播概率到概率亲密度矩阵的社区挖掘算法

摘要 There is growing concern about Community Mining Algorithms based on local optimum as the structure of modern network becomes more and more complex and the scale of it becomes larger and larger.These algorithms have quick calculating speed but a lower accuracy result.In response to these issues,this article proposes a probability matrix of affinity which use Information Propagation to instead of Structural Similarity to represent the global information in the network.This method improves the CONCLUDE algorithm and at the same time reduces the computational complexity of time from O(d(v)2V) to O(d-(v)V).When compared with other algorithms on several networks,it shows that this algorithm not only improves the NMI on LFR networks but also the modularity on real networks.The result proves our algorithms has higher computational efficiency and precision