Research On Influence Maximization Of Social Network

With the continuous expansion of social networks,the use of the relationships existing in social networks to disseminate information has attracted widespread attention from researchers.Inspired by "Word-of-mouth marketing" and "viral marketing" in marketing,a new research direction has gradually evolved-the problem of influence maximization(the abbreviation IMP problem).In the field of computer science,information communication,sociology and other areas caused a research boom.The research of the IMP needs to analyze the propagation characteristics of information according to different network topologies.Establish a specific information propagation model,and design an efficient algorithm of influence maximization.The purpose of the algorithm is to find a set of the most influential users,(node set)to trigger the maximum spread of information in the network.The research on the maximization of social network influence has practical significance.It has been used in link prediction,recommendation system,network campaign,public opinion alert,and emergency notification.In the field,as the scale of the network expands,the problem of maximizing influence becomes increasingly valuable.It can help us to solve problems such as the dissemination of important information in large-scale networks.At the same time,the existing influence maximization algorithm has the following problem.It is difficult to reduce the running time while ensuring the accuracy of the propagation algorithm.In order to solve the shortcomings of the existing work,this dissertation proposes new research ideas.The main work of this dissertation includes follow contents.(1)This dissertation gives a brief introduction to the basic concepts of social networks.Based on this,it introduces the definition of influence maximization problems and introduces two basic influence propagation models,introduces the algorithm evaluation indicators of the maximization of influence,and finally introduces there are two types of existing maximization algorithms,and conclude that there are insufficient existing algorithms and the points that will be improved in this dissertation.(2)This dissertation proposes a structural hole based influence maximization algorithm(SHBG).SHBG algorithm first calculates the structural hole values for all the nodes in the network,sorts the nodes according to the structure hole values,selects candidate node sets from the sorting nodes according to a certain proportion,uses a static greedy algorithm to optimize the candidate nodes,and finally obtains a seed node set.Since the algorithm selects seed nodes only from a small number of structural hole candidates,it saves running time and guarantees propagation accuracy.The experimental results show that the SHBG algorithm can effectively reduce the running time while ensuring the impact propagation accuracy.(3)This dissertation proposes a degree and clustering coefficient based influence maximization algorithm(DCBG).DCBG algorithm combines the two indicators of node degree and clustering coefficient,ranks the importance of nodes in the network,selects candidate nodes from the sorting nodes according to a certain proportion,uses static greedy algorithm to optimize the candidate nodes,and finally obtains a seed node set.The experimental results show that the DCBG algorithm can approximate the propagation accuracy of the greedy algorithm and effectively reduce the running time.Compared with the heuristic algorithm,the propagation accuracy is effectively improved.