Research And Application Of Overlapping Community Detection Alogorithm Based On Modularity Optimization

In recent years,big data and artificial intelligence have continued to develop and become more and more efficient,which have been widely adopted in daily life.The network carries huge and complex data information.The study of its internal structure containing a wealth of information undoubtedly appeals for the birth of techniques in community structure detection strongly.Analyzing the structure of the network and describing the characteristics of the network can understand the characteristics of the network better,which can be used to predict the law of network replacement.Among the traditional community detection methods,non-overlapping community detection algorithms have been studied more.However,with the deepening procedure of research,overlapping community detection methods can match the actual network better,opening up new research ideas for us.Based on the SLPA algorithm and modularity optimization,an improved overlapping community detection algorithm MLPA+ is proposed in this dissertation.The modularity optimization is integrated with the SLPA algorithm to reduce the randomness in SLPA when spreading labels.The maximum square root of the modularity ratio is displayed to make the algorithm divide the initial community in the early stage.In the label propagation,the optimal neighbor node is chosen by comparing the indicator of Adamic-Adar.When there are multiple optimal neighbor nodes,the label with the most occurrences among their labels is selected with updating the current node,so that the result of community detection is more in line with the real community structure.Through the community detection on five sets of artificial network data and four sets of real network data,comparative analysis of experimental results showsthat the advantages of MLPA+ algorithm far outweigh the traditional methods.In addition,the convergence of the algorithm is faster,and the result is more stable than before.It is concluded that the algorithm proposed in this dissertation has better stability and robustness.This algorithm is involved in a cooperation projects "application software of community detection in a specific industry" conducted by a municipal police department to provide effective help in the process of case detection and more effective analysis methods in determining the sensitive groups of organizations involved in activities.The MLPA+ algorithm is used to detect overlapping nodes accurately,narrow the range of sensitive people,and provide an important auxiliary role in case detection.The main improvement and merits of MLPA+ are determining overlapping nodes,avoiding the low accuracy of non-overlapping algorithm,and providing more accurate information for criminal investigation to speed up the progress of case handling.