面向服务生态网络的服务可信性理论研究

The services ecological network is an ecological network, in which the business services are the main part, and which closer integrating with the social networks and physical networks. The services interaction and coordination must be built on the basis of trust. The project aims at the characteristics of the services ecological network, such as cross-domain, cross-network, open, dynamic, self-evolution, and focus on the credibility support for services. We will research the credibility of the service description method based on ontology and fuzzy logic, and depicting the formation mechanism of the trust between services in the service ecological network, the evolution mechanism of services credibility, and establish a complete representation system and evaluation method for service credibility. Based on Bayesian network, we will realize the credibility transfer and polymerization for polymorphism and cross-domain service; Improve LDA model based on the maximum entropy so as to realize service topic detection, and fulfill the service subnet detection based on weighted directed network diagram, then combining the two methods to realize the credible services recommendation in large-scale service network; Based on the Fluent Calculus and action reasoning, we will realize trusted cooperation between the cross-networks, cross domain, heterogeneous services; Based on the service context reasoning to percept the service environment, we will realize the dynamic security mechanism for services credibility. Finally, we will use the electronic business service ecological system to fulfill the theoretical proving, and then provide reference and theoretical support for the description, measurement and control of service reputation and credibility in services ecological network.

服务生态网络是以业务服务为主体，与社交网络、物理网络等紧密融合的聚生态网络，网络中的服务交互与协同必须建立在互信的基础上。本项目针对服务生态网络的跨域、跨网络、开放性、动态、自演化等特性，围绕服务可信性支撑需求，研究基于本体和模糊逻辑的服务可信性描述方法，刻画服务生态网络中服务之间的信任形成机理、服务可信性的演化规律，建立基本完备的服务可信性表征指标体系和评估方法。基于贝叶斯网络，实现多态跨域服务可信性传递与聚合；基于最大熵改进LDA模型实现服务主题发现，基于加权有向网络图实现服务子网发现，并结合这两种方法实现大规模服务网络中的可信服务推荐；基于流演算和行动推理实现跨网、跨域、异构服务之间的可信协同；基于服务上下文推理实现服务环境感知的服务系统可信性动态保障机制；最后，我们将通过电子商务服务生态系统进行理论验证，为服务生态网络中信用和可信性描述、度量和控制等提供参考和理论支撑。

以业务服务为主体，与社交网络、物理网络等紧密融合的服务生态网络已成为当前研究的焦点。然而网络中的服务交互与协同必须建立在互信的基础上。本项目针对服务生态网络中的服务信用和可信性理论进行研究，为跨域、跨网、开放、动态、自演化的服务生态网络提供可信性理论支撑，建立多尺度、多维度的服务可信性度量指标方法和形式化描述。在对已有的服务评论模型进行深入分析的基础上，结合服务可信和网络特性，通过构建基于图的综合声誉模型来建立社交商务信任模型；针对服务网络中信任的传递特点，采用基于复合信任的概率矩阵分解模型建立合理的信任网络，连接信任孤岛；通过最大条件概率的单依赖聚合模型，解决服务之间相互独立的假设不足，提升服务子网发现模型性能；基于凸包构建的信任感知算法解决服务的优化选择问题；通过引入外部数据协助评分矩阵分析的方法和潜在空间学习解决服务网络中可信推荐的冷启动和数据稀疏问题；基于标签和评分的协同过滤模型实现服务标签系统中的服务推荐机制；利用服务评论的社交上下文和服务信任网络实现服务系统的可信性预测，解决欺诈过滤问题。.通过四年的攻关研究，我们团队取得了丰硕的成果，发表了14篇SCI期刊论文，1篇EI期刊论文，8篇国际会议论文，获得了1个国家发明专利授权，2个软件著作权，1项浙江省科技进步一等奖。培养了5名博士生，9名硕士生。达到并超过了任务书的预期要求，圆满完成了该基金的研究任务，为后续的研究项目打下了坚实的基础。