基于博弈论的社交媒体分析与理解

Social media analysis and understanding are now facing some key challenges, including the cluttered network structures, the flood of themes, and the increasing improvement of users’ intelligent and cognitive ability. In this project, we plan to tackle these challenges from the following three aspects: (1) we study the influence of user behavior and decision making on the formation of a social network structure, formulate the noisy social network structure formation as a two-stage game, and detect the community structures based on maximum likelihood estimation and game equilibrium; (2) we formulate the interaction between a user and a social media theme as a game, analyze the equilibrium, as well as characterize and predict the popularity dynamics based on the equilibrium; (3) we use the deal selection on Groupon with Yelp data as an example to illustrate the decision learning process in social media, analyze the statistical properties of the Groupon-Yelp data, and formulate the deal selection problem as a dynamic statistical game as well as derive the optimal selection strategy using the multi-dimension Markov decision process. The main objective of this project is to introduce users’ decision making and interactions into the social media analysis and understanding, explore the corresponding scientific laws, and thus build the theoretical foundation for efficient social media analysis and understanding.

针对社交媒体分析与理解现在面临的网络结构混乱、主题泛滥、用户智慧度和认知能力逐步提高等现象，本项目拟开展如下研究：（1）研究用户行为和决策对社交网络结构形成过程的影响机理，对噪声社交网络结构进行两步博弈建模，基于博弈均衡利用最大似然估计检测社区结构；（2）对用户与社交媒体主题交互进行博弈建模，分析博弈均衡并基于均衡对流行度动态过程进行描述和预测；(3) 以基于Yelp数据的Groupon打折商品选择为例子来研究社交媒体中的决策学习，分析Groupon-Yelp关联数据集统计特性，并以此为基础进行动态随机博弈建模，利用多维马尔可夫决策过程推导最优选择策略。本项目的总体目标是基于博弈论将用户决策和互动引进社交媒体分析与理解，探索其中存在的科学规律，研究相关的基础理论与关键技术，为实现高效的社交媒体分析与理解奠定基础。

针对社交媒体分析与理解现在面临的网络结构混乱、主题泛滥、用户智慧度和认知能力逐步提高等现象，本项目开展了如下研究：（1）研究了用户行为和决策对社交网络结构形成过程的影响机理，对噪声社交网络结构进行两步博弈建模，基于博弈均衡利用最大似然估计检测社区结构；（2）对用户与社交媒体主题交互进行博弈建模，分析了博弈均衡并基于均衡对流行度动态过程进行描述和预测；(3) 基于Yelp数据的Groupon打折商品选择研究了社交媒体中的决策学习，分析了Groupon-Yelp关联数据集统计特性，并以此为基础进行动态随机博弈建模，利用多维马尔可夫决策过程推导了最优选择策略。本项目基于博弈论将用户决策和互动引进了社交媒体分析与理解，探索了其中存在的科学规律，研究了相关的基础理论与关键技术，为实现高效的社交媒体分析与理解奠定基础。