【作者】 徐新卫；

【导师】 丁秋林；

【作者基本信息】 南京航空航天大学 ， 计算机应用技术， 2007， 博士

【摘要】 Web和基于Internet的应用系统改变着当今世界,Internet从信息发布平台逐渐演变为一个开放的分布计算环境,越来越多的数据资源、计算资源与应用资源依托Internet和Web成为可被公共获取和访问的网络资源。近年来,随着“服务”成为开放网络环境下资源封装与抽象的核心概念,通过Web服务实现计算资源灵活配置成为技术发展的创新思路,特别是随着主动计算特征出现和服务要求以人为本、并增加用户体验性,Web主动服务己成为面向服务计算的核心内容。本文以主动式Web服务为基本思想,将用户、服务、业务有机结合,通过实现Web主动服务以促进企业内部以及企业之间的信息集成和交互。本论文的工作主要包括以下几个方面。(1)提出了Web主动服务模型根据Web主动服务架构和主动计算特征,提出了Web主动服务“推”-“拉”模型(WASM),在此模型中对服务请求者的意图进行发现与辨识,以期望得到用户的服务需求,服务质量和等级;同时在服务注册库中查询发现服务;通过服务获取技术绑定所需服务的URI以获取匹配的服务。(2)提出了一种预模式可生长神经网络模型和基于混合策略的服务挖掘算法根据反映用户意图的内外模式和认知过程中的模板匹配、认知变焦原理,提出了一种预模式可生长神经网络,它具有预模式的选取和随着辨识要求自适应增加网络节点(或)和增加模式网络协同工作,可以高效准确地发现和辨识用户意图,并以模式网络的最优输出作为教师样本来训练其它模式网络,以达到知识共享、提高辨识效率和精度;并根据用户意图在主动服务中的不同作用层次提出基于混合挖掘算法有针对性地进行意图的发现与辨识:用v-SVC的支持向量学习算法对用户的偏好和意图进行学习和辨识,反映用户的状态和所处的角色;用概念层次生成算法对用户服务属性层次化,获得用户的服务焦点;借助于模糊聚类算法来辨识用户所期望的服务质量或等级。(3)提出了服务命中策略和小世界模型下的Web智能路由算法考虑在大量冗余、相近的服务信息中和分布式的服务注册库中高效发现所需服务,提出了服务命中策略并通过相近度计算获取相近的可替换服务,解决高效服务命中和服务绑定失效的问题,并根据服务网库的分布独立性,提出基于资源最优的分布式服务网库搜索算法。还依据小世界网络具有特征路径短、聚类度大的特征和服务聚集性,将提供Web服务的对等节点构造成为具有小世界属性的网络,在服务查询和路由过程中可以提高效率和减少通信量。Web服务路由凭借其虚名称和路由感知机制,在SOAP中方便地实现路由路径的指定和装载。借助于基本蚁群算法,提出了在小世界网络模型中基于QoS约束下的智能路由算法。(4)提出了基于反射中间件的服务失效处理方法服务失效是影响服务高可用的制约因素,对于Web主动服务就需要对服务进行必要的动态配置和优化。利用反射中间件技术,将其配置在服务器和客户端捕捉到服务失效的状态,来调整和配置服务以适应服务的内外部环节的变化和要求,并利用混合神经网络作为反射中间件的检测机制。(5)提出了一种基于QoS上下文反馈控制机制的服务组合方法QoS是Web服务关键因素,从基于QoS的服务发现机制和基于QoS的服务合成机制基础上提出了QoS指标的上下文模型,并依据此模型来利用反馈控制机理,使用遗传算法作为控制器对QoS上下文的各项指标进行调节,并在备选的服务中发现所需最优的服务。更多还原

【Abstract】 The Communication and interaction have rapidly changed by Web Technology and application system based on Internet. The Internet is now undergoing an evolution from a simple platform for information sharing to an open distributed computing infrastructure. With more and more data resources, computing resources and applications accessible via the Internet and web pages for public usage, resource sharing and application integration across organizational boundaries. In recent years, the concept services in open of network has been proposed as a mean of abstracting and wrapping environment, which makes Web service the natural way for dynamic resource aggregation and flexible application integration. Especially, along with the features of active computing, human centered and enhancing customer experience in services, which make Web active services become a kernel on SOC (Service Oriented Computing).Implementing Web active services makes possible for information integration and interaction across organizational boundaries and enterprise interior depend upon unite clients, services, and operation together. The main contributions of this dissertation are summarized as fellows:(1) The WASM (Web Active Services Model) that has“push”and“draw”two ways character is proposed.Based on framework of active services and features of actives computing, the key technologies and implementation methods in the WASM are presented. Firstly, discovery and recognizing the asker of services to know its requirement, intention, QoS, and level of services; Secondly, Querying and discovering services information in registry of services; finally, binding proper services’URI to get it that meet user needs.(2) A PENN and Based on hybrids policies algorithm are discussedThe PENN (Patterned Extendable Neural Network) is presented for discovering and recognizing user intention by selecting pro-pattern and adjusting number of latent layer node and/or increases pattern cooperation to meet discovering and recognizing needs. Taking an optimum pattern output as teacher train other patterns for sharing knowledge and improving efficiency and precision. The hybrid algorithm is proposed based on the different activities of user intention in Web initiative services. The preference learning with v -SVC provide effective preference recognition and role selection for user intention. The service focus point will be found with concept hierarchy algorithm, which conceptualizes service attributes. Using fuzzy clustering for user desire level or quality of services is feasible decision.(3) The policy of services hitting and Based on small World model Web routing algorithm are proposedHow to discover required services in abundant redundant and similar service that record in databases which possibly have distributed features? The policy of services hitting is introduced, which effectively solves problem of services hitting, and process invalidation of services binding through similarity degree computing and sorting. Because of distributing and independency of services net database, a resource optimization search algorithm is proposed. Becaude the small world networks have a short length of characteristic path and evident attention of clustering, the efficiency of communication is improved in Web services querying and routing on the network which has small-world features and is constructed by some peer to peer services that regard as vertex in graph. Web service depends on virtual name and routing aware mechanism in expediently implement routing appoint and load in SOAP. Based on basic ant colony algorithm, the QoS constrained Ant colony Algorithm is presented.(4) The mechanism to avoid Web services unavailability is proposedWeb services’high usability is restrained by Services unavailability. The reflection technologies are presented to improving Web services self-adaptive and robust ability to modify and optimize to Web services. The reflection layers are deployed in client’s and server’s sides to catch the traps that are some status of Web services unavailability. The reflection layers can adjust the internal structures and states of services, which avoid services unavailability in time and make services adapt to environment changing. The Hierarchy mixture of expert neural network as a detector in reflective middleware is introduced.(5) The method of Services composition based on QoS mode is proposedOne of key factors is QoS for obtaining services. The QoS context is presented to deal with the different circumstance in compositing services and discovering services. Based on QoS context model, the candidate services are selected and compounded by QoS control system in which the genetic algorithm as controller to ensure to select suitable web services’QoS in the feedback system.更多还原