Research On Intelligent Optimization Algorithm And Application For Influence Maximization

With the advent of the 5G ear,applications such as online office,online meetings,webcasting,and online classrooms have become normal in life.It has completely changed people's traditional work styles and lifestyles,that is,people's production and lifestyles have changed from offline behavior in the traditional sense to online and offline integrated behaviors.In addition,the individual intersection relationship network abstracted based on people's production and life behavior can intuitively portray people's various interactive behaviors relationships.In reality,individuals have influences on their neighbors based on emotions and cognition in the process of interaction.Research on this influence based on the spread and evaluation of network influence will reveal the network dynamics of people's production and life behaviors.The effective control and utilization of this network dynamics is of great practical significance to the improvement of production efficiency and the reduction of management cost.The problem of Influence Maximization is an important research topic in network science,which essentially includes two aspects of research content.First,based on a certain evaluation mechanism to identify a set of nodes with importance and influence.Secondly,based on a certain optimization strategy,the overlapping effect of the influence of nodes in this group is eliminated,so that it can maximize the spread of information or influence in the network based on a certain propagation model.At present,scholars in different fields have put forward many different methods from different angles to solve the problem of influence maximization.With the continuous expansion of network scale,these algorithms have problems such as low efficiency and weak scalability in practical applications.In recent years,intelligent optimization algorithms have achieved good results in the recognition application of social network influence maximization node sets.However,this kind of algorithm based on different search strategies and search mechanisms in discrete network space has some problems,such as poor stability,easy to fall into local optimal solution,and the efficiency of solution needs to be further improved.Therefore,this dissertation studies the solution and application of intelligent optimization algorithm in the problem of influence maximization,and verifies its effectiveness in two kinds of real networks.The main research work of this dissertation is as follows:(1)After analyzing the defect of discrete particle swarm optimization(DPSO),which is easy to fall into the global suboptimal solution due to precocity,a strategy to enhance the local search ability of DPSO algorithm by using the centrality of node near neighborhood is proposed.Through experiments on the real social network Data Sets,it is found that there is a saturation effect when enhancing the local search ability of the DPSO algorithm based on the degree centrality of the nearest neighbor nodes.Specifically,the performance of the algorithm is improved most significantly based on the centrality of the node degree of the 3-hop near neighborhood,and the wider the neighborhood,the algorithm performance decreases.Based on this phenomenon,an improved DPSO algorithm DPSO?NDC based on the 3-hop near-neighborhood node degree centrality local search strategy is proposed.The experiments in six real networks shows that the performance of the proposed algorithm is better than the original algorithm.(2)Although the Discrete Bat Algorithm(DBA)has the advantage of high efficiency in solving the problem of the influence maximization of social networks,it inherits the selection evolution strategy based on random number in the basic bat algorithm,which makes the algorithm have the characteristics of weak convergence stability in the finite iterative evolution process.The strategy of constructing candidate seed node pool based on network clique structure was proposed to enhance the diversity and pertinence of the selection of location vector nodes in the evolutionary process of bat population,so as to improve the stability of the algorithm in the convergence process.Based on this,a clique?DBA algorithm is proposed.Through experiments in several real networks,the effectiveness of the algorithm's performance and convergence stability are verified.(3)When a single intelligent optimization algorithm is used to solve the problem of influence maximizing in the large-scale network,there is bottleneck in efficiency.Based on the multi-core processor technology,the parallel intelligent optimization algorithm is used to identify the most influential node set in large-scale networks,which is an effective way to solve the problem of efficiency bottleneck.In this dissertation,the crow search algorithm is re-encoded and re-designed for the network structure to realize parallel computing,and a discrete crow search algorithm(DCSA)based on parallel search computing is proposed.Experimental results on six large-scale real networks show that the proposed algorithm not only has competitive performance compared with other state-of-the-art algorithms,but also significantly improves its efficiency,which is suitable for parallelizing the problem of influence maximization in large-scale network structures.(4)This dissertation abstracts the network model from two different application scenarios,analyzes the network characteristics and identifies the node set with maximum influence,respectively.(I)The dynamic behavior network model of Linux kernel is constructed,and the network characteristics of the software network are revealed.According to the actual application scenarios,the corresponding influence evaluation model is proposed,and the intelligent optimization algorithm is used to identify the node set with maximum influence.The effectiveness of the intelligent optimization algorithm is evaluated from the aspects of function coverage and overlap ratio with other central algorithms.(II)A book borrowing behavior network model of university library is constructed,some network characteristics of this network are analyzed,and the practical significance of these characteristics are explained in the dissertation.Based on the realistic background,the influence evaluation model of the node is proposed,the intelligent optimization algorithm is used to identify the influence node set,and the effectiveness of the intelligent optimization algorithm is analyzed and applied in the book reading promotion service.The research results of this dissertation show that the intelligent optimization algorithm has higher efficiency and performance when solving the problem of influence maximization in large-scale networks.Based on the corresponding evaluation model,the experiment and analysis of the intelligent optimization algorithm in the real network with different structural characteristics verify the feasibility and effectiveness of the intelligent optimization algorithm in identifying the node set with influence maximization in the real networks.