【作者】 俞东进；

【导师】 王光明；

【作者基本信息】 浙江工商大学 ， 企业管理学， 2010， 博士

【摘要】 现代信息技术的发展及其广泛应用正试图支持、同时也推动着企业管理变革。信息化不仅直接影响了企业的组织结构变革和业务流程再造,而且还重新塑造和整合了企业与其协作伙伴之间的关系。随着企业规模的持续扩大和市场竞争的日益加剧,业务的复杂性不断攀升。作为企业管理的核心：决策,在当今信息化时代,至少面临着如下三个方面的挑战：1)决策环境日益复杂。决策需要跨越系统的边界、部门的边界甚至企业的边界。2)决策要求不断提高。为了在竞争中胜出,决策过程必须更为快速,必须适应动态变化的业务需求,同时决策成本却要保持不变甚至尽量降低。3)决策资源更加丰富。海量数据和遗留系统的普遍存在,使得决策的难点从数据的获取转向了信息的处理。决策支持系统(Decision Support Systems, DSS)利用计算机和软件等技术手段,在人们制定决策的过程中提供辅助支持,以帮助做出有效的决策。作为信息化的一个重要内容,企业决策支持系统的构建和应用也在支持和推动企业管理变革的过程中起到了关键性的作用。但是,传统的决策支持系统,无论是三部件结构、三系统结构、四库三功能结构,往往忽视决策需求和决策目标不确定的客观特征,部件之间的耦合性较强,内外接口缺乏标准化,普遍不具备协同处理、重组织和柔性扩展能力。同时,对传统决策支持系统的扩展需要引入DSS生成器,不但成本较高,而且扩展幅度较小。而上世纪90年代开始兴起的面向服务计算(Service Oriented Computing, SOC)技术,可以解决传统决策支持系统的上述问题。这里,与服务管理领域定义的服务是一种顾客作为共同生产者、随时间消逝的、无形的经历不同,在面向服务计算领域中,服务是特指那些能实现某项具体业务功能的、可被复用和重组、遵循标准化接口协议的基本软件组成单元。本文采用面向服务计算技术对传统决策支持系统进行了重构,即通过对各类异构的分布式业务组件以标准化服务的形式进行封装以及对服务的动态自由组合,形成一个能在复杂环境中实现快速动态重构和扩展的开放式系统。基于服务的决策支持系统强调系统整体框架的松耦合,具有可扩展和柔性自组织的特征,对内可以通过对组成决策支持系统的各个部件以服务的方式进行动态选择、适配、组装和调用,对外可通过与相关系统的信息集成和应用耦合实现与分布式企业计算环境的融合,从而赋予了决策者在集成化的综合环境下选择的灵活性和随企业战略动态调整而快速变化的能力,同时也降低了重构和扩展的成本。本文首先给出了基于服务的决策支持系统(SBDSS)整体框架的层次化概念模型和实现模型。SBDSS的整体框架分为交互层、服务组合层、服务层、业务组件层和资源层5个层次。特别地,服务层包含了分布于企业计算环境中的可被复用和重组、遵循标准化接口协议的基本软件组成单元；服务组合层可以根据需要选取服务层提供的不同的原子服务,并通过顺序、循环、选择、并行等不同的组合方式编排出具有实际意义的求解方法。SBDSS采用服务组件架构(SCA)实现服务建模,采用服务数据对象(SDO)作为数据和消息模型,采用业务过程执行语言(BPEL)用于编排服务,采用基于JBI的企业服务总线技术完成系统组合。论文在提出上述基于服务的决策支持系统整体框架的基础上,进一步研究了基于服务的数据管理问题。在如何为DSS提供提供一致的、全局的数据视图方面,论文提出了基于消息的多数据集同步整合模型(MEDSIM)。MEDSIM采用标记置信度的方式解决了来自于不同数据源的主数据冲突和业务数据冲突。同时,针对复杂决策环境下数据分布、异构、海量等特点,论文提出了一个基于XMLA协议的分层多维数据分析框架(xDAF)。客户端程序可以向xDAF发起基于SOAP协议包装的XMLA请求,并获得相关的XMLA结果响应。为了改善大规模海量数据的多维查询性能,xDAF采用了基于对象软引用技术的聚合池对象置换策略和实视图技术,实验证实相关方法是合理和有效的。基于为决策者提供友好、灵活的决策环境才是辅助决策者处理复杂问题的最佳策略这一论点出发,论文进一步提出了面向决策支持系统的柔性化层次模型。该柔性化层次模型可理解为一个逐步演进的层次化结构,分为适应的柔性、重构的柔性、服务的柔性和集成的柔性等4个层次。论文特别研究了重构柔性和服务柔性的主要实现途径。在如何实现重构柔性方面,提出了一个支持快速重构的通用DSS框架。该框架实现了专用DSS期望的默认行为的类集合,专用DSS则可通过扩展通用DSS框架的子类来支持相关的专有行为。在如何实现服务内部柔性方面,提出了基于控制反转的设计思路,并通过SCA和Spring的组合模型,实现了服务实例在基于服务的决策支持系统整体框架中的内外统一接入。论文最后给出了基于服务的劳动力(人力资源)市场决策支持系统的原型设计和应用,同时验证了上述理论的正确性。该系统采用Apache Tuscany作为面向服务架构的运行时(Runtime)平台,采用Apache ServiceMix作为ESB容器。系统提供了分布式环境中面向劳动力(人力资源)市场的多主题通用多维分析引擎以及基于分类、聚类和关联的通用数据挖掘算法库,可视化地支持决策过程的人机交互,可实现地域劳动力(人力资源)需求和供给的特征分析、结构分析、转移分析、趋势分析,以便为劳动力(人力资源)市场的相关决策提供技术支持。系统既可为各级政府劳动力市场宏观决策提供支持,其部分模块也可适用于大中型企业人力资源规划。论文围绕基于服务的决策支持系统的基本特征、框架结构和柔性化机制,取得了如下创新性成果：(1)提出了SBDSS整体框架的层次模型、概念模型和实现模型。(2)提出了SBDSS环境中基于消息的多数据集同步整合模型：MEDSIM。MEDSIM采用标记置信度的方式解决了来自于不同数据源的主数据冲突和业务数据冲突。(3)提出了SBDSS环境下基于XMLA协议的分层多维数据分析框架：xDAF。xDAF同时采用基于对象软引用技术的聚合池对象置换策略和实视图技术,以改善对海量多维数据的查询性能。(4)给出了SBDSS的柔性化层次模型,特别是提出了通过通用DSS规范化框架以达到重构柔性、通过基于控制反转机制和SCA／Spring组合模型以达到服务柔性的实现机理。论文所进行的对基于服务的决策支持系统的研究具有重要的现实意义。它可应用于构建新一代企业决策支持平台,帮助企业决策者灵活、便捷、快速地适应动态变化的决策环境和要求,帮助企业决策者获取更多更丰富的、跨部门(甚至跨企业边界)的决策资源,从而支持和推动企业管理变革。更多还原

【Abstract】 The development and its widespread application of modern information technology have been trying to support, and meanwhile drive enterprise management reform. Informatization not only impacts the enterprise structure reform and the business process reengineering, but also rebuilds and integrates the relationship between the enterprise and its partners.Due to the fierce market competition, the scale of organizations keeps expanding while the business complexity grows. During the information era, the decision making, as the key of enterprise management, faces at least three challenges.1) The increasingly complicated decision environment. Decisions shoud be made across the boundaries of systems, departments and even enterprises.2) The ceaselessly improved decision demands. Decisions should be made more quickly and adapt to the dynamic changing business requirements with no higher cost.3) The more and more abundant decision resources. The pervasive existent large amount of data and legacy systems turn the difficult points of decision making from data acquiring to information processing.The decision support system (DSS) aids decision makers through computers and software. As one of the main topics of informatization, the enterprise decision support system helps support and drive enterprise reform. However, no matter how it is composed, the traditional decision support systems, with the tight components coupling and non-standard interfaces, often neglect the common sense that both requirements and targets of decision are sometimes indeterminate. They are not capable of flexible decision with the lack of capabilities such as interoperation, reorganization and extendibility. Traditionally, the DSS generator needs to be introduced while extending the specific DSS, which costs greatly but only brings about limited extensions.The appearance of Service Oriented Computing (SOC) technology in 1990s could settle the problems of rigidity and closeness of traditional DSS. Here, different with the concept in the domain of Service Management, the service in SOC denotes the basic reusable software component with specific business functions and standard interfaces. This paper reconstructs traditional DSS architecture using SOC, and moreover proposes the Service based Decision Support System, i.e. SBDSS. In SBDSS, distributed and heterogeneous business components are encapsulated and provided in the form of services. With the loosely components coupling, SBDSS could dynamically select, adapt, compose and invoke the inner service components, and therefore be easily integrated into the distributed enterprise computing environment through the way of information integration and application coupling with reduced cost.This paper first presents the hierarchical conceptual model and implementation model for the framework of SBDSS, which is divided into the following five layers: the interaction layer, the service composing layer, the service layer, the component layer and the resource layer. The service layer contains many fundamental, standard and replaceable software units. In the service composition layer, different atomic services could be selected from the service layer and then be arranged to produce the problem resolving services. As for the implementation model, SBDSS adopts Service Component Architecture for service modeling, the Service Data Object for data and message modeling, the Business Process Execution Language for service assembly respectively, and finally integrates the different units with the help of JBI based Enterprise Service Bus.The paper further studies the issues related with the data management under the service context. It introduces the Message Based Data Synchronization and Integration Model which resolves the data conflicts by setting and comparing the predefined confidence values for different data sources of both the Master Data and the Business Data. It also gives the multidimensional data analysis framework based on the XMLA protocol, called xDAF. The client program could issue the XMLA requests over SOAP to xDAF and receive the responses from xDAF. With the introduction of soft reference of objects in aggregation pools and materialized views, xDAF boosts its performance for multidimensional data query greatly, which is proved by the related experiments.Since the friendly and flexible decision making environment is regarded as one of the best facilities which could be provided to the decision maker, the paper then presents the flexibility model of DSS which consists 4 layers as the flexibility of adaption, the flexibility of reconstruction, the flexibility of service and the flexibility of integration. The paper pays the particular attention to the approach of reconstruction flexibility and the service flexibility. The general DSS framework is proposed to support the rapid reconstruction. It realizes the default behavior for the specific DSS, which could be then extended if necessary. For the service flexibility, the mechanism of inversion of control is proposed to help service instances flexibly composed and easily integrated into SBDSS with the aid of the assembly model of SCA and Spring Framework.Finally the paper introduces the implementation of prototype for the service based labor market decision support system and verifies the above proposed theory. The system employs Apache Tuscany as the SOA runtime and Apache ServiceMix as the ESB container. It provides the multi-dimensional analysis engine for labor market themes and the general data mining algorithms of classification, clustering and association to conduct the feature analysis, the structural analysis, the transferring analysis and the trend analysis of the demand and supply to support decision making in the administration of regional labor resources. The system could be introduced to provide the decision support for labor administration. In addition, some parts of the system could also be used to help enterprise human resource planning.This paper achieves the following innovative points through the study of service based DSS, especially focusing on its fundamental features, framework structure and flexibility mechanism.(1) It presents the hierarchical model, conceptual model and implementation model for the framework of Service Base Decision Support System.(2) It introduces the Message Based Data Synchronization and Integration Model which resolves the data conflicts by setting and comparing the predefined confidence values for different data sources of both the Master Data and the Business Data.(3) It gives the multidimensional data analysis framework based on the XMLA protocol, called xDAF. With the introduction of soft reference of objects in aggregation pools and materialized views, xDAF boosts its performance for multidimensional data query.(4) It presents the layered DSS flexibility model and pays the particular attention to the approach of the reconstruction flexibility with the predefined general DSS framework, and the approach of the service flexibility with the mechanism of inversion of control and the assembly model of SCA and Spring Framework.The study of service based decision support system has significant practical values. Its achievements could be adopted to construct the next generation of decision support system, which helps the decision makers adapt to the dynamic changing decision environments and demands in the flexible, agile and fast manner, and access more abundant decision resources across department and enterprise boundaries. It could then eventually support and drive the enterprise management reform.更多还原