Dynamic Network Community Detection Based On Improved Mayfly Algorithm

With the continuous development of information technology,the network is accessible in daily life,and it will inadvertently produce information interaction,thus forming the network.These network structures are complex,the amount of information is enormous,and it is not easy to find practical information.Community detection can quickly find the community structure of complex networks,which helps essential mine information from the complex information network.Therefore,community detection is of great significance in basic research and engineering applications and has become a vital topic for complex network research.In recent years,with the in-depth study of community detection,the shortcomings of static community detection in network dynamic detection are gradually exposed,and dynamic community detection has been derived.The research on dynamic community detection is essential to improve the shortcomings of static community detection.Aiming at the problem of dynamic network community detection,this thesis introduces the mayfly algorithm and proposes two methods combined with different strategies,namely,a two-stage discrete mayfly algorithm based on evolutionary population(EP-TSDMA)and a multi-objective discrete mayfly algorithm based on boundary point local search(BLS-MODMA).The EP-TSDMA algorithm developed a two-stage discrete mayfly algorithm first.In the first stage,the mayfly algorithm was discretized,and the mayfly individuals after each update position were taken out.The crossover and mutation operations suitable for dynamic community detection were introduced in the second stage.Secondly,the worst mayfly individuals and the best mayfly individuals are adjusted,and the community merging operation is performed on the worst three mayfly individuals to remove the influence of scattered communities on the quality of mayfly,and the point migration operation is performed on the best three mayfly individuals to avoid the wrong division of some nodes.Finally,the evolutionary population strategy is used to reduce the population size and improve search efficiency.In order to increase the diversity of initial solutions,the BLS-MODMA algorithm uses two different algorithms to initialize male and female mayfly populations.Then the position update strategy is modified,and the crossover and mutation operations are redefined to accelerate the evolution process.Finally,the local search is carried out on the boundary points of the optimal global solution to make the optimal solution escape from the optimal local solution and enhance the efficiency of the algorithm.Through experiments on nine datasets and comparison with typical algorithms and novel algorithms in community detection in dynamic networks,the results prove that the EP-TSDMA algorithm and BLS-MODMA algorithm proposed in this paper can effectively identify the community structure in dynamic networks.