基于聚类的复杂网络社团结构发现

Community structure detecting of complex networks based on clustering is proposed because number of community is need to be set in advance and overlapping community structure detection has become research focus in recent years. Fuzzy c-means clustering algorithm is used for community structure detecting in complex network, in which the overlapping community structure is found. The swarm intelligence based fuzzy c-mean algorithm for community detecting is formed by combining swarm intelligence algorithm and fuzzy c-mean algorithm, which avoids the problem of sinking into local extreme. Moreover, the problem of setting initial clustering center is resolved by this algorithm. Space constraints information is added to the traditional fuzzy c-means algorithm and the accuracy of community detecting is enhanced by improving the similarity measure method. Speed of convergence of fuzzy c-means algorithm based on space constraints for community detecting is accelerated by amending the membership function. Mean shift algorithm is used for community detecting, in which the problem of setting the number of community is resolved. Moreover, the accuracy of community detecting is further improved. Finally, the proposed algorithms are used to predict public opinion information network transmission of mass emergency, in which the change of public opinion information network transmission of mass emergency is observed by community structure detecting in complex network and the spread of public sentiment can be controled effectively.

针对重叠结构社团发现已成为近几年研究的热点和进行复杂网络社团发现需要设置社团个数的问题，提出基于聚类的复杂网络社团发现算法。将模糊C均值聚类算法用于复杂网络中社团结构的发现，用于发现复杂网络中重叠的社团结构。将智能群算法与模糊C均值算法结合，形成基于智能群算法的模糊C均值社团发现算法，解决模糊C均值算法易陷入局部极值和需要设置初始聚类中心的问题。将空间约束信息加入到传统的模糊C均值算法，通过相似性度量方法的改进提高模糊C均值算法用于社团发现的准确性。通过修正基于空间约束的模糊聚类社团发现算法中隶属度函数的值，加快基于空间约束的模糊聚类社团发现方法的收敛速度。利用均值漂移算法进行社团发现，解决社团个数的设置问题，进一步提高社团发现的准确度。最后将所提算法用于突发群体性事件网络舆情信息传播的预测，通过复杂网络社团结构的发现观察突发性群体事件舆情信息传播的变化情况，从而有效地控制舆情信息的传播。

本项目针对重叠结构社团发现已成为近几年研究的热点和进行复杂网络社团发现需要设置社团个数的问题开展研究。将基于聚类的智能优化算法用于社团结构的发现。利用粒子群算法和微分进化算法全局性和鲁棒性的特点，提出基于智能群算法的模糊C均复杂网络社团结构发现算法，避免传统模糊C均值算法陷入局部极值。并且通过智能群算法解决传统模糊C均值算法需要确定初始分类数和初始聚类中心的问题。将细菌趋药性算法用于模糊C均值社团结构发现，提出了基于混合细菌趋药性算法的模糊C均值社团结构发现和基于群体细菌趋药性算法的模糊C均值社团结构发现。完成了基于空间约束的模糊聚类社团结构发现算法，提出快速的基于空间约束的模糊聚类社团发现算法，通过对隶属度值的修正，提高算法的速度。将均值漂移算法用于复杂网络的社团发现，解决进行社团发现时需要事先设置社团个数的问题。并力图将所提算法用于突发群体性事件网络舆情信息传播的预测，通过复杂网络社团结构的发现观察突发性群体事件舆情信息传播的变化情况，从而有效地控制舆情信息的传播。