Bayesian Modeling Of Continuous Behavior Of Users In Online Social Networks

As the online social network service industry matures,the growth of registered users tends to be flat.The secondary development of social platform must depend on customer retention rather than customer acquisition.The most concerned problem in customer retention phase is the user's continuous usage or repeat purchase behavior.Thus,this dissertation focuses on a core research object of "continuous behavior of social network users".Starting from the three different perspectives including customer lifetime value theory,social influence theory,and dynamic learning theory,we build three hierarchical Bayesian models upon "information disclosure and dissemination behaviors by users in free open social media platform","repeat purchase behaviors by players in freemium-based online game communities",and "content creation and content consumption behaviors by users in content montization based social networking sites".Those series of empirical studies gradually explore the nature,the driving forces,and the dynamic formation process of individual's continuous behavior,which provides a comprehensive and systematic analysis framework.Firstly,the basic characteristics of individual's continuous behavior are analyzed in Study I,in which historical behavior data are used to capture individual's behavior inertia and regularity so as to predict the future behavior.In terms of theoretical contribution,customer lifetime value theory is introduced to the study of user's continuous usage behavior,in which the empirical modeling is conduced based on individual's real behaviors.As to methodological contribution,stochastic processes and probability models are used to describe the basic stochastic features of individual's continuous behavior and the heterogeneity across individuals,which solves the problems of zero inflation and over-dispersion in traditional models for count data.Secondly,Study II focuses on the external drivers of individual's continuous behavior,especially the interdependent behaviors among individuals embedded in social networks.In terms of theoretical contribution,we extend the research object of innovation diffusion study from binary adoption behavior to repeat purchase at the post-adoption stage,and the inter-correlated structure of continuous behavior among multiple individuals is studied according to the social influence theory.As to methodological contribution,a network auto-regressive Poisson model is constructed to identify and estimate the interaction intensity among multiple individuals,solving the endogeneity problem of social influence identification.Finally,Study ? decomposes the continuous behavior of individuals into a dynamic learning process based on both self experience and external signals.In the context of two-sided market,the dynamic evolvement of continuous behaviors by individuals with different roles(i.e.consumers and suppliers)is studied.In terms of theoretical contribution,it supplements the lack of literature on consumer cognitive learning study in the field of two-side market and models the learning process of content supplier's topic choice decisions.As to methodological contribution,we leverage Bayesian learning models to illustrate the dynamic sequential updating mechanism of individual's continuous behavior and examine the impacts of market signials and market mechanism on individual's behaviors from the perspective of content consumers and suppliers in the two-side market.In summary,this dissertation conducts progressive researches based on three aspects:the basic characteristics of individual's continuous behavior,the interdependence of continuous behavior among multi-agents,and the dynamic evolvement of continuous behavior by individuals with different roles.Our work fills the research gaps in the fields of consumer behavior research,innovation diffusion research,and consumer cognitive learning research.The comprehensive analysis framework on social network users' continuous behavior in this dissertation has profound therotical and methodological impacts on literature.The practical insights will also contribute to the healthy and sustainable development of social media platform.