Analysis Of User Influence In Social Networks

With the development of social networks, people can send and receiveinformation quickly, which leds to the increasing influence of online-life to offline-life. Social marketing has appeared, one of its very important requirement ismaximizing the influence of marketing activities in social networks. Therefore, theanalysis of user influence in a social network become a critical research. It isdesigned to analyze the size of influence given to others, and select the subset ofusers whice have most contribution to the spread of influence to make the greatestimpact ultimately on the network,under the premise of a limited budget.Many sociologists have previously researched this problem in sociologicalview, with the appearence of Facebook, Twitter, and other social networks, moreand more scholars began to study the user influence in a such complex and hugenetwork.There even are a lot of companies which committed to analysis the userinfluence. But,at present,the analysis of user influence is mainly focused onqualitative analysis,few researchers conducted a quantitative analysis.In thispaper,we will introduct the models which has been used in sociology anddissemination of information to sina weibo,and make a detailed comparison ofdifferent methods, on the basis of previous work.Firstly, this paper inspects the methods based on network structure, includingthe Maximum Degree algorithm, the Distance Centrality algorithm which are bothused generally in sociology, and PeopleRank algorithm which is similar toPageRank, and compares those methods with the user’s actual average repost rate.Secondly, we introduce two probabilistic model-Independent Cascade modeland the Linear Threshold model. Algorithm based on these two models not onlytakes the network structure into account, but also focuses on the converse behaviorwith other users in history. The practical result proves that, these two algorithms arebetter than the algorithm which only considered the network structure. For theproblem of maximizing the influence of optimal K subset,the greedy selectionalgorithm based on two models are a approximation not less than63percent tooptimal solution. We add some optimization in order to make two algorithms to bemore efficiently, the actual results show that the optimized greedy algorithmscombined with network structure improves efficiency at the same time does not losetoo much calculation accuracy.Finally, we carried out a analysis of various models and methods, establishedthe concept of―the influence of the structure‖and―the influence of the behavior‖toexamine the user influence from two aspects. We have created a system ofcomprehensive analysis of the influence,which can provide supports in social marketing.