Font Size: a A A

Research On Personalized Influence Maximization Of Social Network

Posted on:2017-07-25Degree:MasterType:Thesis
Country:ChinaCandidate:X WangFull Text:PDF
GTID:2348330488472328Subject:Computer technology
Abstract/Summary:PDF Full Text Request
With the formation the Mobile Internet era,Online Social Networks are developing at a rapid pace these years.It changed the way of modern humans getting information and communication.The new social approach is a more intuitive reflection of the stable social circle which formed in people's real life.At the same time,some of the characteristics that distinguish it from traditional social approach are found.What is more,as number of users explosive growth,the Potential commercial value of social network has attracted attention from all walks of life.Personalized influence maximization problem in social network has become a new branch of influence maximization problem in recent years.The problem is defined as,given a target user,finding the most influential user set for the target which contains less than k users.At present,some well-known social network data analysis company,like klout,peerindex,is using the influence maximization scheme to provide reference for advertising and marketing.As the role of personal values begins to show in the social media age,to seek effective personalized influence maximization solution is the new requirements of social network data analysts.In this paper,the key technology of personalized influence maximization in social network have been in-depth discussion and research.A new algorithm was proposed based on independent cascade model.Furthermore,from the perspective of influence propagation model,a new train of thought was found.The specific work we have done for this paper is as follows:(1)While different from existing research works that assume equal propagating strengths of social network edges,Maximized-Influence-Path Algorithm(MIPA)was proposed based on the independent cascade model to find out the top-k most influential nodes for the target user ignoring the inappropriate assumption.The algorithm use maximized influence paths to estimate influenced strengths of target.Firstly,to compute the propagating strengths from the nodes of social network to the neighbors of the target node,the strengths of edges were transformed into its logarithmic form and the optimized Dijkstra algorithm has been used for getting the maximized influence paths.Secondly,the strength of paths which pass through different neighbors with the same source node have been consolidated to the node's propagating strength to the target.Finally,Seed set was found by selecting the top-k nodes with high propagating strength.(2)In order to escape the inertia that existing research mainly focused on the basal linear threshold model and the independent cascade model,the heat diffusion model is introduced into the research of personalized the influence maximization problem.Heat diffusion process was used to simulate the transmission of the influence between the social network users.To open up new ideas for personalized influence maximization,target heat greedy algorithm been proposed based on the analysis of the heat propagation rules.(3)By using c + + programming language to implement the proposed MaximizedInfluence-Path Algorithm and target heat greedy algorithm,We testified the algorithms based on the real social network data sets.The personalized influence maximization experiments were took place on independent cascade model and heat diffusion model respectively.The groups experimental results validate the performance of the proposed algorithms.
Keywords/Search Tags:social network, personalization, influence maximization, Algorithm, target user
PDF Full Text Request
Related items