Research On The Mechanism Of Social Spreading Based On The Characteristics Of Local Nodes

People's daily life is undergoing profound changes from flourishing social network which also brings opportunities and challenges never seen in the history.Social information dissemination,brightly characterized by high efficiency and universality,has been proved that it is conducive to the forming emerging business models then generating incredible commercial value.Study on the law of information dissemination can help to spread more accurately(like product recommendation and news push)and then promote social economy to prosper.The old and repeated information born in copying,forwarding and endless marketing is prone to make users lose taste for that.As a result,content recommended in compliance with personal preference will facilitate the efficient information dissemination.Additionally,an in-depth probe into the path of social communication will conduce to position the superior partners and boost information dissemination.Therefore,it is of vital significance to explore potential social relationships.As the evolution laws of social information dissemination become increasingly complex,it has been long delved into and addressed in the study on social dissemination:how to identify the biggest potato of communications,filter out redundancies to promote information dissemination and smell out potential social relationships,so as to grasp the pivotal technologies influencing information dissemination and the general evolution trend of information dissemination.Therefore,backed by the research ideas and methods of complex network,recommendation system theory and link prediction theory,this thesis acts pursuant to from the analyzing social phenomena to mining the nature of communication,makes analysis on macro network structure and explore micro social nodes as well as the factors affecting information dissemination,then formulates effective strategies and algorithm models accordingly.The main work and innovative points of this thesis are as follows:1.From a macroscopic view,to study the mechanism of promoting social communication and probe into the influence of widening the communication range and shrinking the outbreak threshold.The results show that selection strategy comparison between overall-situation influence communicator and partial-situation influence communicator presents that the latter one can effectively simplify the algorithm complexity and improve the communication efficiency.Study on the identification strategy of most influential partial-situation communicators puts forward propagation threshold model on the basis of random walk seed selection.An in-depth analysis and comparison on the propagation scale under various selection strategies has been made through the numerical simulation of artificial network and four real networks and then it is concluded that the explosion threshold declines and the propagation scale burgeons when partial-situation Hub nodes are preferred as seed disseminators.The analysis results also secure a solid theoretical basis for the study on social network information dissemination.2.From a microscopic view,to stand on a foot the social preference characteristics of nodes and study the communication mechanism of enhancing information diversity.This thesis,focused on the impact of partial-situation information recommendation on dissemination,reduces the redundant relevance of object attributes in existing recommendation algorithms,and further makes the accuracy and diversity of recommendation a leapfrog.Targeting at the correlation and repeatability of attribute redundancy in information recommendation,this thesis proposes an ERJD-CB algorithm,that means,a detailed analysis on the similarity redundancy of attributes born in objects deduces the minimum deviation value of similarity estimation and then deletes second-order similarity redundancy.At the same time,based on the theory of balanced mass diffusion and heat conduction,this thesis achieves balance compensation of recommendation accuracy and diversity.It can raise the accuracy,novelty and diversity of recommendation while filtering redundant information at the same time.Experimental statistics showcase that ERD-CB model overmatches the existing five mainstream benchmark algorithms in accuracy,novelty and diversity.3.From a microscopic view,to stand on a foot in the social relationship characteristics of nodes and probe into the dissemination mechanism of mining potential paths.This thesis,centering on the prediction of quasi-local propagation paths,presents an analysis result that the quasi-local algorithm can improve the accuracy of link prediction and reduce the implementation complexity,compared with the partial-situation algorithm with low accuracy and the overall-situation algorithm with high complexity.Targeting at the-issues of mining potential propagation paths,link prediction technology can give effective prediction on propagation paths.It is found that nodes can effectively attract unconnected endpoints through the influence of neighbors.Consequentially,the difference of node neighbor influence will bring the difference of neighbor contribution influence to the node.Therefore,this thesis designs a link prediction model based on neighbor contribution(NC).Simulation results on 12 real networks show that the model based on neighbor contribution can effectively improve the accuracy of link prediction in social networks.4.Backed by the research on neighbor contribution,this thesis finds that H index can effectively present the depth of neighbor influence,with regard to the issue that traditional link prediction algorithm only takes the breadth of neighbor influence into consideration but ignores the depth.Through analysis on the connectivity breadth and depth of neighbor networks,this thesis proposes a link prediction model based on neighbor degree and neighbor H index influence(ION-D,ION-H).The experimental simulation concludes that the prediction model,with higher accuracy,effectively elevates the accuracy of social information propagation path prediction,compared with the mainstream indexes.