基于社交影响力分析的个性化推荐方法研究

Recently, the way of modelling and applying social influence has become a research hotspot. However, existing social influence models are too general and could not describe the influence propagation process in the real-world scenarios precisely. Meanwhile, the current works which focus on exploring social influence for recommendation seldom consider the personalized requirements from social users or service providers. To this end, with the help of social influence models and recommender systems, this project studies the key issues on personalized recommendations based on the analysis of social influence, and these issues are related to social influence, social users and service providers. Along this line, we will first propose a model to study the context-aware information transmission probabilities, and thus the principle of interactions among network structure, information content and user contexts could be revealed. On this basis, we then design the strategy of maximizing the given user’s influence by recommending social links, and propose the seed user recommendation method to help service providers maximize the awareness of their information where some personalized constraints could be predefined. We hope that this could improve the satisfaction of the users and increase the profit of the service providers. Finally, the above discoveries will be further applied to the scenario of analyzing the group-effect to enhance the performance of traditional recommender systems. In summary, by exploring the mechanism of information propagation, this project will propose effective solutions for the problem of personalized recommendations based on the analysis of social influence. In this way, not only could the application areas of the social influence analysis be expanded but also the performance of the recommendation technologies could be enhanced. Thus, the research of this project has important theoretical and practical value.

社交影响力的建模分析与应用已成为一个热点研究方向。然而，已有的影响力泛化模型很难真实地反映现实世界中的信息传播过程。而且，当前利用社交影响力进行信息服务的研究，较少考虑用户和商家在获取与传播信息时的个性化需求。针对上述问题，本项目以社交网络中的信息、用户和商家为对象，以影响力模型和推荐系统为手段，开展基于社交影响力分析的个性化信息推荐关键技术研究。首先，研究情境感知的信息传播概率评估模型，揭示网络结构、信息内容与用户情境间的相互作用原理。在此基础上，设计面向个体用户影响力最大化的社交邻居推荐策略，以及基于商家个性化营销约束的种子用户推荐方法，从而增加用户满意度与商家收益。最后，在推荐系统的群组效应分析问题中检验上述研究成果。本项目通过探索信息传播机制，提出了基于社交影响力的个性化信息推荐有效解决方案，拓展了影响力分析的应用领域，提升了个性化推荐方法的实际效果，具有重要理论意义和应用价值。

对社交影响力进行建模分析研究具有重要理论意义和应用价值。然而，已有的影响力泛化模型很难真实地反映现实世界中的信息传播过程。而且，当前利用社交影响力进行信息服务的研究，较少考虑用户和商家在获取与传播信息时的个性化需求。针对上述问题，本项目以社交网络中的信息、用户和商家为对象，以影响力模型和推荐系统为手段，开展基于社交影响力分析的个性化信息推荐关键技术研究。首先，研究情境感知的信息传播概率评估模型，揭示网络结构、信息内容与用户情境间的相互作用原理。在此基础上，设计面向个体用户影响力最大化的社交邻居推荐策略，以及基于商家个性化营销约束的种子用户推荐方法，从而增加用户满意度与商家收益。最后，在推荐系统的群组效应分析问题中检验上述研究成果。此外，还在用户的社会化行为建模、社交用户画像、基础方法研究等主要方面取得了创新性研究成果。在本项目的支持下，已发表学术论文30篇（IEEE/ACM Trans系列5篇，IJCAI、KDD、AAAI等CCF推荐的A/B类会议15篇），申请发明专利1项，并与腾讯、微软、科大讯飞等公司建立了密切合作关系。本项目通过探索信息传播机制，提出了基于社交影响力的个性化信息推荐有效解决方案，拓展了影响力分析的应用领域， 提升了个性化推荐和口碑营销方法的实际效果。