Research On Influence Maximization Algorithm Based On Intelligence Optimization

In the past few decades of rapid popularization of Internet information technology in all walks of life,various social instant messaging software applications have played an important role in life services.These communication services have involved all aspects of people's lives.The interaction and communication between people have become more and more convenient and intensive,and online social networks have gradually evolved into tools and carriers for instant information release and dissemination.The popularity of online social networks has drawn widespread attention from all social groups to the spread of information,as a message can quickly spread through the entire social circle through the social relationship among online friends.Different from the traditional information propagation media,the advantage of online social network as an information communication carrier is that it is not limited by time and space and has a very wide range of influences.Therefore,information diffusion in network social network attracts extensive research in computer science,physics,epidemiology and other fields.At present,the analysis of social complex network is a hot topic in the academic circle,and the influence maximization of social network is an important branch of its research field.The problem of influence maximization is to select a small number of users in a complex social network,who have the greatest influence spread under a specific propagation model.The greedy Monte Carlo simulation approach theoretically guarantees a near-optimal solution,but it is very inefficient.Although many heuristic methods have been developed,they greatly reduce the quality of the solution.The current challenge of influence maximization research is how to better balance the influence spread and algorithm running time in complex and large social networks.In order to solve this problem,this article focuses on the combination of the construction of seed node candidate pool and swarm intelligence optimization algorithm.This dissertation includes:1.Firstly,this paper systematically expounds the definition of the problem of influence maximization and its related theoretical concepts,investigates the research work of influence maximization at home and abroad,and analyzes the shortcomings of the current research work.Some representative classic influence maximization algorithm will be introduced in detail.2.Through a lot of research on the influence maximization problem,the problem can be formally defined as a discrete combinatorial optimization problem,and the heuristic intelligent optimization algorithms have obvious advantages in dealing with complex optimization problems.Therefore,this paper proposes an influence maximization algorithm based on harmony search(HSIM).The algorithm utilizes nodes with higher structure similarity score to initialize the harmony memory,and uses a second-order average degree heuristic method to construct candidate seeds to speed up the algorithm.Experiments on four academic cooperative network data sets show that the HSIM algorithm has high efficiency and the influence spread of seed nodes is ranked in the first tier.3.Currently,most algorithms run slowly because the influence spread of the node set uses tens of thousands of Monte Carlo simulations.This paper presents a local probabilistic solutions strategy which is similar to Monte Carlo simulation to calculate the influence spread of a node set,and proposes an influence maximization algorithm based on immune genetic algorithm(IGIM).The node set influence spread of this strategy requires only one accurate calculation,which can effectively avoid tens of thousands of simulation calculations of node-set's influence propagation.Through experiments on four data sets of academic cooperative networks,results demonstrate the efficiency and accuracy of the proposed algorithm in solving the influence maximization problem.In terms of influence spread,it has extremely similar performance with the current best performing CELF algorithm,and the running time is about 5 orders of magnitude faster than CELF algorithm.