Research On Theory And Applications Of Social Influence Propagation

With the rapid development of Internet technology and social networks,there are hundreds of millions of people in social networks which have served as important platforms for people to obtain and share information.In this context,the propagation of social influence has received tremendous attention.The basic mission of studying social influence theory is to deeply analyze human interaction mode and information propagation mode by mining human behaviors,the structure of social networks,and other attributes.By studying the theory of social influence propagation,a group of influential users in social networks can be selected to promote products or services,The process can be modeled by influence maximization problem,which aims to select a 6)-size subset of users(usually called seeds)with maximum influence propagation in social networks.Since social influence incorporates human interaction and information propagation,it has been widely applied in numerous real-world application scenarios such as viral marketing,social recommendation,human behavior prediction,and so on.Social influence analysis based on social network structure and applications of social influence in traditional real-world scenarios has been widely studied.However,the effect of human mobility on social influence propagation and applications of social influence in emerging fields need to be further studied.Therefore,it is important to analyze social influence propagation based on users’ spatio-temporal behaviors and expand applications of social influence in emerging fields such as Spatial Crowdsourcing(SC).Consequently,this dissertation summarizes three key problems of social influence analysis and its application,i.e.,the role of users’ spatio-temporal behaviors on social influence propagation,the effect of communities on social influence analysis,and social influence-aware task assignment in Spatial Crowdsourcing.The main conclusions are as follows:1.Similarity-aware influence maximization: Most of the existing studies on influence maximization in social networks only consider the structure of social networks or the location information to analyze the influence propagation of users.However,the approaches can not fully reflect the effect of users’ spatio-temporal behaviors on influence propagation.To overcome these limitations,this dissertation analyzes users’ behavior pattern from the time aspect and the space aspect,and then users’ spatio-temporal similarity is derived.Moreover,a propagation to consumption model is designed to analyze the effect of users’ spatio-temporal behaviors on influence propagation.To solve the similarityaware influence maximization,the dissertation proposes an influence propagation trees based algorithm and a cutting tails algorithm.The extensive experiments over real-world datasets demonstrate the efficiency and effectiveness of the proposed algorithms.2.Community-based influence maximization: Due to the large scale of social networks,it is relatively complicated to calculate the influence propagation of users.A variety of heuristic algorithms are proposed to reduce the time complexity of influence propagation calculation.However,most of the heuristic algorithms fail to guarantee the accuracy of computing influence propagation.In order to synthesize both the high accuracy and low time complexity of influence propagation calculation,this dissertation proposes a community-based influence maximization problem.Specifically,a weighted distance algorithm which considers the users’ spatio-temporal behavior and the structure of networks is proposed to accurately detect communities.Based on communities,the dissertation proposes a variety of influence maximization algorithms,which ensure high-accuracy influence propagation computation and reduce the time cost of selecting seeds.3.Influence-aware task assignment: Most of the studies on applications of social influence focus on traditional fields such as viral marketing and social recommendation.In an emerging field,Spatial Crowdsourcing,most of the existing studies exploit social network structure to improve the quality of task assignment,while ignoring users’ interaction mode and information propagation mode.To overcome the limitations,this dissertation proposes an influence-aware task assignment problem.Users’ social influence,users’ affinity towards tasks and locations of tasks are taken into account when assigning tasks.A variety of assignment algorithms are proposed to solve the task assignment problem,and extensive experiments on real-world datasets offer detailed insight into the effectiveness of the proposed solutions.The proposed framework can provide an effective solution when task requesters need to propagate the information of tasks in social networks.