基于Web Services的群体决策支持系统研究

【作者】 赵阳；

【导师】 肖亮；

【作者基本信息】 南京理工大学 ， 模式识别与智能系统， 2008， 硕士

【摘要】 群体决策算法和群体决策支持系统的研究是系统工程以及人工智能领域的热点问题,在复杂系统评估、大型工程项目决策分析等实践中具有广泛的应用前景。现阶段群体决策支持系统对非结构化、半结构化决策问题还存在定性知识推理手段欠缺,定性数据向定量数据转化困难,软件系统复用性和跨平台能力不足等缺点。本文以构建基于Web Services的群体决策支持系统为研究主线,以航天项目风险评估为应用背景,主要集中研究了半结构化和非结构化问题的模型算法,最后搭建了WIFGDSS原型系统。本文研究成果包括:(1)对于非结构化问题,将粗糙集理论以及定性推理理论引入GDSS中,为系统提供智力支持。对于粗糙集,在经典理论基础上将其在群体环境下进行推广,解决了定性知识获取的问题;对于定性推理的QSIM算法,提出了用未确知有理数描述定性知识,用群体决策方法改善不确定参数关系建模,将物理推理的QSIM与专家经验知识结合,提高了推理的创造力。(2)对于半结构化问题,提出了双参数偏差模糊层次分析法。对于现有偏差矩阵法,指出了其中参数选取的错误,提出了更为精准的双参数计算模型及其约束条件,从实验数据给出了参数的取值规律,并通过航天风险实例与现有的直接法、偏差矩阵法做比较,验证了改进的有效性。(3)为了提高系统复用性和跨平台能力,搭建了一个基于Web Services的原型系统WIFGDSS。系统融合了Web服务的思想,用服务封装各个算法模型,实现了粗糙集推理、QSIM定性推理、模糊层次群决策和Web服务管理四大功能,使系统从传统C/S转化为B/S架构;搭建了基于Private UDDI的分布式模型库,扩充了模型覆盖范围,减少了开发成本。本文阐述了系统三大层次四个模块的实现,对核心的决策模块和模型库模块进行了分析,体现了系统良好地决策支持和高效地模型管理这两大优势。最后将本文架构的WIFGDSS原型系统应用于航天项目风险评估。实践证明,系统具有较好的群体推理和模糊层次分析能力,也具有较好的系统复用性和跨平台性。更多还原

【Abstract】 Study on algorithms of Group Decision and Group Decision Support System (GDSS) are hot issues in the field of Systems Engineering and Artificial Intelligence. They are widely applied in complex system assessments, decision-making for large-sized projects and so on.There are some deficiencies in the existing GDSS as encountering Non-Structure and Semi-Structure problems. For example, lack of efficicent method to reason the qualitative knowledge, difficulty in transforming qualitative data to quantitative data, insufficiency in reusage of software and usage of different operating systems. This dissertation takes the construction of GDSS based on Web Services as the major line. Using the risk evaluation of aerospace projects as the background of application, it mainly studies on the algorithms for Non-Structure and Semi-Structure problems and builds a practical system WIFGDSS (Web Services based Intelligent FAHP GDSS). The main works of this dissertation are as follows:(1) For non-structure problem, this dissertation bring Rough Set theory and Qualitative Simulative for Imprise Model (QSIM) into GDSS in order to support the system in an intelligence aspect.The classical Rough Set is expended under a group enviorment, which could enrich the acquirement of qualitative knowledge. QSIM has been improved in three ways: using the unknown rational number to describe qualitative knowledge, applying the theory of group decision to modeling the uncertain relationship of the parameters, integrating the physical logic-based qualitative reasoning method with experience and knowledge of experts-based reasoning method, which enhance the creativity of the QSIM.(2) For semi-structure problem, this paper proposes a new algorithm of double-parameter -Fuzzy AHP using deviation matrix (DPDM-FAHP). By proposing a double parameter based expression, it improves the deficiency of construction of fuzzy consistent weight matrix in existing FAHP using deviation matrix. The constraints and the values rule of parameters also has been given by analyzing results of simulation experiment. Applied to the risk evaluation of aerospace projects, it is proved that DPDM-FAHP has a better performance than the existing FAHP methods by comparing the result with theirs.(3) In order to improve the reusage of the software and the usage of different operating systems, this dissertation builds a practical system WIFGDSS based on Web Services. WIFGDSS is built by assimilating the views of Web Services.There are two huge differences between WIFGDSS and other GDSS.One is that the each algorithm exists as a service, which makes the architecture of the system changed from C/S to B/S. And the services implement the four function of the system: Rough Set Reasoning, QSIM qualitative reasoning, Group Decision with FAHP and Web Services Management. The other one is that building up a distribute modle base adopting Private UDDI. It expends the range of coverage of distribute modles, and reduces the cost as well. This dissertation specifys the implementation of the big three layers and the four modules. By analyzing the two core modules, which are decision module and modle base module, the two advantages of WIFGDSS has been showed: excellence support to decision-making and managing modles with high efficiency.Finally, WIFGDSS is applied to the Aerospace Project Risk Evaluation. The practice proves that this system possesses better group reasoning ability and fuzzy-hierarchy Analytic ability.At the mean time, it could be reused rapidly and accessed to different operating systems.更多还原