Research On Positive Influence Maximization In Social Network

With the popularization of the Internet and the rapid development of information technology,more and more people use social networks for daily communication and information acquisition,and social networks have become an indispensable part of people’s life.Information spreads rapidly in social networks,trending topics on social platforms are likely to happen only a few minutes ago,and information spreads widely,theoretically covering the entire network.As new users continue to join,the scale of social networks is getting bigger and richer in content and data.Based on the above characteristics,many businesses begin to viral marketing in social networks,to promote and popularize their new products,and the influence maximization problem is the most important issue in viral marketing,it aims to find k seed nodes in social networks,these nodes can achieve maximum’s sphere of influence in the social network.In recent years,many scholars have conducted in-depth research on this issue and obtained some valuable research results.However,the current research on influence maximization is almost only carried out in unsigned networks,ignoring the polarity relationship between users.In certain scenarios,the required range of influence can be overestimated if polarity is not taken into account.Among polar-related influence maximization problems,the most useful one is positive influence maximization.For example,in a viral marketing scenario,merchants focus more on the word of mouth range,or positive impact range,of the product.Therefore,this paper studies the maximization of positive influence in signed network,and the main research work is as follows:(1)Based on the idea of reverse influence sampling,a positive influence maximization algorithm RIS-S was proposed.Firstly,the algorithm considered the polarity relation between nodes when generating the reverse reachable set,and filtered the nodes that can be added to the reverse reachable set,which improves the quality of the reverse reachable set.Then,the maximum sampling depth was limited because nodes overlap between reversely reachable sets.Finally,a comparative experiment was conducted in three signed network data sets.The experimental results show that RIS-S algorithm performs better than IMM algorithm of the same type in solving the problem of positive influence maximization.(2)The problem of positive influence maximization is NP-hard.The traditional greedy algorithm and heuristic algorithm have their own shortcomings in solving the problem.Therefore,this paper proposes a positive influence maximization algorithm GA-PIM based on genetic algorithm.In order to improve the quality of the initial population,GA-PIM algorithm selected the individuals of the initial population through the validity screening strategy in the initial population initialization stage.In the mutation stage,the selection of replacement nodes was extended to the whole signed network to avoid the algorithm falling into the local optimal value.Experimental results show that the accuracy of GA-PIM algorithm is close to that of the Greedy algorithm IC-P Greedy,but the running time is greatly reduced,so it can be considered that the algorithm has reached a good balance in terms of accuracy and efficiency.