【作者】 李亚飞；

【导师】 刘业政；

【作者基本信息】 合肥工业大学 ， 管理科学与工程， 2004， 硕士

【摘要】 证据理论是处理由认识的局限性所带来的不确定性问题的有力工具，它处理的证据来源于专家。但专家的知识经验往往是有限的，获取也较困难，且可能存在一定的主观性。粗糙集理论反映了人们以不完全信息或知识去处理一些不可分辨现象的能力，或依据观察、度量得到某些不精确的结果而进行数据分类的能力。粗糙集的提出为处理模糊信息系统或不确定性问题提供了一种新型数学工具，是对其它处理不确定性问题理论如概率理论、证据理论、模糊集理论等的一种补充。针对上述所提证据理论的局限性，本文提出了一种基于粗糙集理论的证据获取的新方法，并对证据合成和应用进行了研究。 首先对粗糙集理论作了进一步的研究，细化了其划分的粒度，在此基础上，对决策表的决策属性作了进一步的转换，结合粗糙集和证据理论之间的关系，再利用粗糙集的分类思想和隶属度概念，计算证据的基本可信度分配，从而实现了证据信息的提取。 其次对证据合成的现状做了研究，针对目前方法在证据相关性、冲突性、重要性等方面无法很好解决的问题，从以解决证据独立性问题为目的出发，给出了采用属性聚类的方法分解决策表来解决问题的方案，并对属性约简的现状和问题作了进一步的研究，设计了一种更有效的属性约简算法。 再次，讨论了证据重要度和支持度的概念，研究了基于本文所提理论的证据的合成方法以及决策支持。 最后，在上述研究内容的基础上，本文研究了基于粗糙集证据理论的股票分析预测系统的体系结构、设计与实现等，并通过试验验证了本文所提方法的正确性。更多还原

【Abstract】 As a powerful tool in dealing with uncertainty questions, the evidence using in the Evidence theory is given by experts. But the expert’s knowledge is limited, subjective and difficult to gam sometimes. With the Rough Sets theory, people can process the undistinguished problem using uncompleted information or knowledge. And also enhance the capability to classify the imprecise data coming from observation or measurement. So the presentation of the Rough Sets theory gives us a new mathematic tool processing the fuzzy information system and uncertainty problem, and is a valuable complement of theory such as Probability theory, Evidence theory, Fuzzy Sets theory etc as well. To the limitation of Evidence theory mentioned above, this paper proposes a new way of knowledge acquirement and also presents a valuable method of the evidence combination and application.Firstly, through further research of the Rough Sets theory, we refine the granule of the classification, and then change the decision attribute of the decision table. Base on these and the relationship between Rough Sets theory and Evidence theory, we compute the evidence’s basic belief assignment through classification, and thus we realize the acquisition of evidence.Secondly, we discuss the current research of the evidence combination. To the limitation that the current methods of the evidence combination can’t well solved such as interrelation, conflict and importance, and for the goal of the solving the independence of the evidence, a new solution that using the attribute cluster to partition the decision table is proposed. In this paper we also discuss the current research of attribute reduction and give a new method as well.Thirdly, we present two new conceptions: evidence importance and evidence support. And then propose the evidence combination and decision support under the theory this paper present.Finally, on the basis of the efforts above, the design, architecture and realization of the stock market analysis and forecast system base on Rough Sets-based Evidence theory is proposed. And through the experiment we validate the availability of the theory.更多还原