Research On Nodes Influence Maximization Modelling And Community Detection Methods In Social Networks

Social networks are an analytical perspective for studying the real world.The nodes influence maximization and community detection have become current research hotspots and key issues.A large number of social,commercial,and management applications have made the study of these issues very urgent.These studies can help to successfully advertise e-commerce products,optimize the use of limited marketing budget resources,and assist in the prediction and intervention of epidemic immunization,public opinion monitoring,and network transmission of emergencies and group events.Based on the complex networks theory,multi-attribute decision making theory,multi-objective optimization theory and data mining technology,this thesis studies the influence maximization and community discovery in the social networks.The main contributions and innovations of this thesis are listed as follows:(1)Establish a comprehensive evaluation model of node influence in social networks based on multiple indicatorsFrom the four perspectives of local,global,position and propagation dynamics,this thesis selects the centrality indicators and regards it as the attribute of decision criteria,then builds a comprehensive evaluation model of node influence based on multi-attribute decision making theory.We propose a dynamic weighted TOPSIS method for identifying influential nodes in complex networks.Here,based on the grey correlation analysis theory and the susceptible–infected–recovered model,an attribute dynamic weighting strategy is proposed.Finally,the complexity analysis and experimental verification are carried out to prove the effectiveness and practicability of the proposed method.(2)Construct a model for influence maximization of top-k node set based on heuristic clusteringIn this thesis,a heuristic clustering model of top-k node set influence maximization problem is constructed,and an adaptive heuristic clustering algorithm with both efficiency and effectiveness is proposed.There are two core innovations as follows.First,to speed up clustering iteration and avoid local optimization,an initial cluster centers selection strategy based on extended neighborhood coreness and minimum distance is proposed.Secondly,a dynamic local similarity index based on path is designed,which can be dynamically adaptive adjusted to the optimal mode when the average shortest path of a given network is different,so as to achieve higher accuracy.Finally,the complexity analysis and experimental verification analysis are carried out to prove the effectiveness of the method.(3)Build a model of influence maximization under limited budget based on NSGA-II genetic strategyCombined with the practical application of marketing cost-benefit,this thesis studies the problem of influence maximization,and constructs a multi-objective optimization model aiming at influence maximization and cost minimization.Based on the NSGA-II model,this thesis proposes a new algorithm to identify the set of seed nodes with influence maximization under limited budget.In the algorithm,the reduction strategy of searching range of seed set is given,which can effectively reduce the computational complexity on the premise of ensuring the searching effect.Experiments on real social networks verify the effectiveness of the proposed model.(4)Propose a community detection algorithm based on multi-objective genetic evolution strategyFocusing on the research of community detection algorithms based on multiobjective evolutionary strategy,this thesis proposes a multi-objective genetic optimization community detection algorithm based on the classification and the topology information.The algorithm first gives the initial population random probability generation stategy based on node similarity.Then,the strategy of chromosome crossing based on the classification is proposed,so as to strengthen the local exploration of the better individuals and the global interaction of the worse individuals.Finally,a community correction strategy based on topological information is designed.To verify the effectiveness of the algorithm,this thesis conducts comparative experiments with several typical community detection algorithms on GN benchmark network,LFR benchmark network and real social network data sets.