社会化商务环境下的动态信任机制研究

With the emerging of social commerce after traditional E-commerce, user social networks based trust management is one of the key factors to ensure its sound growth. Nowadays, the research of trust mechanism in social commerce is at the initial stage with many unsolved problems, such as rating utility values mapping based on user segmentation, mechanism of reprocessing and propagation of rating information by propagating users, as well as the asynchronous nature of multi-paths rating information and its dynamic combination. Therefore, considering the social calculability of social commerce, the project will focus on 3 main subjects: (1) We propose an user segmentation model based on unified granulation user rating and social attributes. And then, we design a method to compute the utility value of rating information on the basis of user segmentation model to improve its accuracy and personalize; (2) In order to model user network in social commerce, we develop a dynamic time-varying hyper-graph combined with five elements such as time, domain, action, structure and user generated content. Based on the hyper-graph, we construct user trust networks to further investigate the active reprocessing mechanism of user rating information in user trust network, and then obtain the computing methods of rating information’s propagation and evolution; (3) In order to implement dynamic trust combination for asynchronous multi-paths rating information, a new dynamic evidence combination scheme is also studied. The research results of this project will be the theoretical foundation and technical support of trust management in social commerce.

继传统电子商务之后，社会化商务开始兴起，基于社会网络的信任管理是影响其健康发展的一个核心要素。当前，社会化商务中信任机制研究尚处于起步阶段，有诸多问题亟待解决，如评价效用值与用户细分类型的映射、传播用户对评价信息的再加工与传播机制、评价信息多路径传播的异步性与动态信任合成等。因此，本项目从社会化商务的社会可计算性出发，重点研究：(1)构建基于统一粒度评价信息与社会属性的用户细分模型，设计用户细分情形下的评价效用值计算方法，提高评价信息效用值计算的准确性；(2)研究融合时间、领域、行为、结构、内容五要素的时变超图模型来建模社会化商务中的用户网络，并由此构建用户信任网络，进而研究信任网络中的传播用户对评价信息的主动再加工机制，获取评价信息的传播与演变机理；(3)构建一种动态证据合成方法，实现多路径异步评价信息的动态信任合成。本项目研究成果将为社会化商务中的信任管理提供理论支撑和技术保障。

针对社会化商务中信任管理的个性化、动态性和准确性需求，聚焦用户属性、用户评论/评分细粒度信息获取与动态分析；评论和信任信息传播以及多源异步时间下的信任计算等研究内容，提出了有效解决方法，并将相关研究成果应用到全域智慧旅游系统平台中，在实际的社会化商务环境下取得了较好的效果。提出了用户和商品属性特征自动抽取方法、面向用户评论方面主题的细粒度评论情感分析、基于评论主题的动态评分预测与跨平台用户评分预测、基于主题分析的用户评论聚类方法等，为商品和用户细分及评分值映射提供了解决方案；提出了基于Transformer的信任信息传播预测模型、基于Soft-Voting的集成学习信任传播预测模型，实现了对信任信息在社交网络中传播机制的刻画；基于遗忘机制和情绪感染理论，设计了基于滑动窗口的信任值动态演变计算方法和加权模糊性证据理论合成方法，解决了多源多路径信任效用值的动态合成。最后，将相关研究成果应用到智慧旅游系统中，基于用户社交属性和评论内容计算用户和商品信息的可信性，验证了项目研究成果的有效性。相关研究成果为社会化商务环境下融合时间、领域、行为、结构、内容的动态信任管理提供了有效解决方案。