喻园管理论坛 2018年第69期(总第413期)

演讲主题: Machine Learning for Social Network Analytics

主 讲 人: 方晓教授, 美国特拉华大学商学院

主 持 人: 杨彦武教授,工商管理系

活动时间: 2018年7月6日（周五）上午9:30-11:00

活动地点: 管理学院119室

主讲人简介: 方晓, 美国特拉华大学商学院教授, 摩根大通学者。他同时兼任特拉华大学金融分析研究院、计算机系和电子工程系教授。方晓教授的研究领域包括：机器学习、大数据分析,、社会网络分析和金融科技。方晓教授的研究成果发表于Management Science, Operations Research, MIS Quarterly, Information Systems Research等著名管理学学术期刊。他的研究成果还发表在计算机领域一流学术期刊上，包括 ACM Transactions on Internet Technology, ACM Transactions on Information Systems, IEEE Transactions on Knowledge and Data Engineering。方晓教授的研究成果曾被麻省理工科技评论, 麻省理工史隆管理评论, 伦敦政治经济学院商业评论等媒体报道。他获得2016年国际运筹和管理学会(INFORMS) Design Science Award。方晓教授现为管理信息系统国际顶级刊物MIS Quarterly副主编。

活动简介: Social networks such as those facilitated by social media, online games, or mobile devices have attracted increasing attention from both academia and industry that explore how to leverage such networks for greater business and societal benefits. Toward that end, we develop novel models, theories, and methods that mine massive social network data for business purposes. In this project, we focus on a unique phenomenon in social networks – the diffusion of adoption behavior (e.g., adoption of a product, service, or opinion) from one social entity to another. Specifically, we investigate three critical and related problems concerning this phenomenon: adoption, persuasion, and link recommendation. That is, the diffusion of adoption behavior is initiated by persuaders and reached to adopters through the linkage structure of a social network. Accordingly, we study the following problems: how to predict adoption probabilities in a social network? how to predict top persuaders in a social network? and how to recommend links for a social network? Let us take the problem of predicting adoption probabilities as an illustration. Adoption probability refers to the probability that a social entity will adopt a product, service, or opinion in the foreseeable future. Building on relevant social network theories, we identify key factors that affect adoption decisions: social influence, structural equivalence, entity similarity, and hidden factors. The principal challenge thus is how to predict adoption probabilities in the presence of hidden factors that are generally unobserved. To address this challenge, we develop a Bayesian learning method on the basis of the expectation-maximization framework. Using data from two large-scale social networks, we demonstrate that the developed method significantly outperforms prevalent existing methods.