【作者】 孙剑；

【导师】 彭秀艳；

【作者基本信息】 哈尔滨工程大学 ， 系统工程， 2019， 博士

【摘要】 基于面向服务的体系架构(Service Oriented Architecture,SOA)的软件开发方式得到了软件行业和学术界的极大关注。Web(World Wide Web)服务是目前作为SOA中服务的一种主要实现方式。在SOA架构下,基于Web服务的软件开发主要包括Web服务选择组合和供应商参与开发方式。通过Web服务选择组合可以实现从纷繁的Web服务中选择最优的服务,将它们组合成为高质量的软件产品。通过供应商参与开发,企业可以借助供应商专业技术能力来参与软件产品的设计,以实现特定要求的软件产品。由于目前软件工程中并没有严格的统一尺度来度量Web服务的各种属性,导致不能简单的利用制造行业中选择标准组件的方法来解决Web服务选择问题。同时对于Web服务选择和供应商参与开发过程中任务指派问题还需要考虑许多约束条件的限制,比如在不确定性环境下Web服务供应商的任务指派、基于QoS(Quality of Service)属性约束下Web服务选择问题等等。因此,研究Web服务选择和供应商任务指派问题具有重要的理论意义和工程价值。目前,对基于Web服务的软件开发多局限在从计算机科学技术的角度进行研究。本文运用系统工程的理论和方法对Web服务选择和供应商任务指派过程中涉及的若干问题进行深入研究。对用户需求、QoS属性(如可靠性、可用性等)和供应商的各种特性(协同能力、开发周期等)进行客观描述,并基于这些定量的描述构建决策模型和方法,为软件工程管理人员提供决策服务。(1)针对QoS属性取值以及用户偏好存在不确定性的问题,提出了一种基于VIKOR-GRA的直觉模糊多属性Web服务选择方法。该方法利用直觉模糊集对QoS属性值和用户偏好进行客观描述,并在此基础上将VIKOR和GRA方法相结合,实现了多属性Web服务的选择排序。结合应用实例并通过灵敏度分析、与直觉模糊TOPSIS-GRA方法的对比验证了本文提出的决策方法具有更好的稳定性、有效性和辨识度。(2)研究了基于工作流的Web服务组合选择问题。针对传统智能算法在求解Web服务组合选择优化问题时存在寻优能力不高、稳定性差和容易陷入局部最优,导致出现早熟的问题,本文提出一种改进的烟花算法,利用灰色关联分析法对烟花算法选择策略进行改进。同时为了提高算法局部挖掘能力和全局开采能力,将高斯变异和柯西变异策略相结合形成混合变异算子,并在混合变异算子中引入邻域结构,进而提高算法局部搜索的性能。(3)研究了基于用户需求的Web服务选择问题。针对传统QFD在确定QoS属性权重过程中对用户满意度和需求竞争性因素考虑不足的问题,本文提出一种基于QFD-IFS、Kano模型和信息熵的QoS属性权重综合确定方法。利用QFD-IFS方法将用户需求转换为QoS属性,计算出QoS属性的基础权重值。引入Kano模型确定用户需求满意度指数,并对基础权重值进行改进,得到满意度权重值。考虑本企业用户需求的实际表现,利用信息熵计算QoS属性竞争性修正因子,从需求竞争性关系角度对满意度权重值进行改进,进而改善了用户需求分析的准确性和客观性。(4)针对供应商开发能力、费用、时间、协同关系和Web服务相关性信息存在不确定的问题,提出了协同开发环境中Web服务供应商的模糊任务指派模型。根据用户偏好程度,通过语言量化OWA算子系数,调节指标权重系数,将复杂的模糊多目标优化问题化简为单目标优化问题来进行求解。实现了协同开发环境中供应商开发的Web服务性能最优、协同开发时间最短以及成本最短的需求目标。更多还原

【Abstract】 The software development methods based on Service-oriented Architecture(SOA)stand out and arouse wide concern among the software industry and academic circles.Web(World Wide Web)services,as a kind of technology applied in distributed computing,are now regarded as a major implementation of services in SOA.The Web services-based software development methods under SOA are mainly divided into two types,namely composition and selection of Web services and supplier involvement in development.Through composition and selection of Web services,the best services can be selected from numerous web services to constitute high-quality software products.Through supplier involvement in development,companies can utilize the expertise of their suppliers to participate in the design of software products and provide specific software products.Since no uniform and stringent scales have been used to measure various attributes of Web services in software engineering,it is impossible to solve the problems in selecting Web services simply by means of the methods that the manufacturing industry adopts to select standard components.Moreover,in the Web services-based software development projects,many constraints,such as task assignment of Web services providers in uncertain environments and selection of Web services based on QoS(Quality of Service)attribute constraints,shall be considered for selection of Web services and task assignment of suppliers involved in development.Therefore,the research on selection of Web services and task assignment of suppliers has great theoretical and engineering significance.Nowadays,Web services-based software development is mostly studied from the perspective of computer science and technology.Operations research and management science are more appropriate perspectives for investigating the selection of Web services and task assignment of suppliers involved in development.Mathematical theories and methods can be used to objectively describe user requirements,QoS attributes(e.g.reliability and availability)and various characteristics of suppliers(interoperability and development cycle).Based on the quantitative descriptions,decision models and methods can be established to provide decision services for software engineering managers.By taking the train of thought above as the way of research and the software development of Web services as the background,several problems in selection of Web services and task assignment of suppliers were investigated in depth in accordance with the theories and methods of system engineering.(1)The value of QoS attributes and the uncertainty of user preferences were investigated.Intuitionistic fuzzy sets were used to objectively describe user preferences and QoS attribute value.On this basis,the VIKOR and GRA methods were combin ed.An intuitionistic fuzzy multi-attribute method based on VIKOR-GRA was proposed for selection of Web services.Based on the instances in selection of Web services,sensitivity analysis and comparison with the intuitionistic fuzzy TOPSIS-GRA method,the proposed method was verified and proven to have better stability,validity and degrees of differentiation.(2)The composition and selection of Web services based on workflow were investigated.When applied to solve the optimization problem in composition and selection of Web services,traditional intelligent algorithms have low optimizing capability and stability and can easily fall into local optimum,thus resulting in premature problems.Therefore,an improved fireworks algorithm was proposed.The gray relative analysis method was used to improve the selection strategy of fireworks algorithm.Meanwhile,to improve the algorithm’s capabilities in local mining and global mining,the strategies for Gaussian mutation and Cauchy mutation were combined to form a mixed mutation operator.The neighborhood structure was also introduced into the mixed mutation operator,so as to improve the algorithm’s performance in local search(3)The selection of Web services based on user requirements was investigated.Given that traditional QFD does not adequately consider user satisfaction and competitiveness of requirements in determining the weight of QoS attributes,a new method was proposed to determine the weight of QoS attributes comprehensively based on QFD-IFS,Kano model and information entropy.The QFD-IFS method was used to convert the user requirements into QoS attributes,and calculate the basic weight value of QoS attributes.The Kano model was introduced to determine the user satisfaction index,improve basic weight value,and finally obtain the weight value of user satisfaction.In view of the actual performance of user requirements in the enterprise,information entropy was used to calculate the competitive correction factor of QoS attributes.The weight value of user satisfaction was improved from the competitive relationship of requirements.As a result,the accuracy and objectivity of user requirements analysis were finally improved.(4)The development capabilities,cost,time and synergic relationship of suppliers and the uncertainties of Web services-related information were investigated.A fuzzy model of task assignment was proposed for Web service providers in collaborative development environment.Language was adjusted to quantify the coefficient of OWA operator and regulate the weight coefficient of index in accordance with the degree of user preference.The complex fuzzy multi-objective optimization problem was reduced to a simpler single-objective optimization problem.With the proposed method,it will take the shortest time for suppliers to provide the best Web service performance at the lowest cost in collaborative development environment.更多还原