【中文摘要】 随着社会的进步和发展,决策信息系统的研究和应用已取得很大的进展,但在信息社会到来的今天,人们所面临的决策问题日趋复杂,大量的、不完全的、有噪声的、模糊的、随机的实际数据干扰着决策者提取隐含的、有价值的决策信息。因此,提供一套解决含有此类不确定信息的决策信息系统一不确定型决策信息系统的决策方法已迫在眉睫。粗糙集理论是二十世纪八十年代初由波兰数学家一Z.Pawlak提出的一种刻画不确定性和不完整性知识的数学工具,这为处理不确定型决策信息系统提供了一条新思路。该理论近年来日益受到广泛关注,己在人工智能、知识发现、故障诊断、模式识、专家系统等方面得到了成功的应用。以粗糙集为工具处理不确定型决策信息系统为我们提供了一条新思路。本文通过对现有复杂决策问题特征的深入分析,在大量检索国内外资料、跟踪国际前沿技术,总结和借鉴前人经验的基础上,将粗糙集理论与决策问题相结合,本文的具体研究内容如下:研究了决策信息系统的数据预处理过程,在噪音数据的有效识别问题上,提出了一种粗糙聚类算法。在属性约简方面,提出了基于同族矩阵的属性约简方法,并研究了同族矩阵的性质。在连续属性的离散化问题上,从聚类相似度以及分类质量...更多、分类精度的角度对经典离散化方法进行了比较,得到了离散化方法相同的优劣序。在不确定决策信息系统的设计上,将粗糙集与神经网络结合,构造了粗糙神经网络决策系统,实证分析表明粗糙神经网络缩短了网络的训练时间,同时分类精度也有了明显提升。针对属性值为区间数的决策系统,提出一种新的离散化方法,并结合灰色系统理论,定义了决策规则,通过对比神经网络,该区间数型粗糙灰色决策系统简化了决策规则,提高了智能决策效率。
【英文摘要】 As social progress and development, research on Decision Information System and its application has made great progress, However, in the information society is coming, the decisioin problems we are facing are becoming more and more complicated, valuable information with large, incomplete, noisy, fuzzy, random data interfere with decision makers seeking interesting and valuable information. So it is urgent to provide a suit of decision methods which can solve the uncertain decision information system.The rough set theory proposed by Pawlak (1982) is established on the basis of database, when database is uncertain or incomplete. It provides one new way for us to solve the uncertain decision information system. The rough set theory has preferable application on Artificial Intelligence and Knowledge Discovery, Pattern Recognition, Fault Detection, Expert Systems, etc.Based on great searching of internal and external inofmration and following closely international advanced technology I deep...ly analyze the characteristic of the complex decision and combine rought sets theory with classical decision methods. The main contents of this paper are as follows:On the data preprocessing of Decision Information System, on account of distinguishing noisy data, one rough cluster algorithm is proposed. In respect of attribute reduction, the redundant set of attributes is obtained by constituting homogenous matrix whose properties are discussed later, whereas that differs with classical method. In terms of discretization, four kinds of typical discretization algorithms were comparatively analyzed from two aspects with examples, one referred to the variable quality of classification and accuracy of approximation under different parameter, the other was the similarity degrees between reducted variable sets and the original variable set, Finally, the consistent conclusion on preference of discretization algorithms were gained.As far as uncertain Decision Information System is concerned, an approach of back propagation neural network with rough set(RSBP) is presented, Simulation results indicate this model,compared with the conventional BP neural network model,can reduce the training time and improve the accuracy of classification.In respect of Information System with interval numbers, a discretization algorithm is proposed. Combing with Grey System Theory, decision rules are defined, Simulation results indicate that Interval Rough—Grey Decision Information System Simplify the decision-making rules and improve the efficiency of intelligent decision-making.
【中文关键词】 粗糙集; 属性约简; 属性离散化; 神经网络; 遗传算法; 灰色关联度
【英文关键词】 Rough set; Attribute reduction; Discretization; Netural network; Genetic algorithm; Grey relational degree
