【作者】 梁凯春；

【导师】 蔡淑琴；

【作者基本信息】 华中科技大学 ， 管理科学与工程， 2007， 博士

【摘要】 企业的决策问题中有很大一部分都是定义不完善、非结构化的劣构问题,这些问题的求解直接关系到企业的生存和发展。劣构问题的问题成分未知,问题空间不明确,存在多种解决方案、解决路径,或者根本没有确定的解决方案,其解决方案是难以预见的或发散的,因此相对于决策者所掌握的问题领域知识来说是复杂的。这类问题的求解一般都是使用头脑风暴法,求解过程中使用到的信息、知识,以及所产生的新知识得不到总结和管理,知识散布在个人头脑中。因此如何有效的管理劣构问题求解过程中的知识,支持知识的交流、互动、共享,辅助劣构问题的求解,成为企业亟需解决的问题。本文针对劣构问题及其求解过程中的知识管理进行了研究,其主要内容如下:文章首先界定劣构问题的概念和范畴,分析劣构问题具有的特征和分类。分析劣构问题求解过程中的各种活动,提出劣构问题求解过程的模型,形式化描述劣构问题和劣构问题求解过程。通过一个企业中劣构问题求解的实例证验,劣构问题求解过程模型可以降低劣构问题求解的复杂度,提高劣构问题求解效率,规范化劣构问题求解过程。论文分析劣构问题求解中知识需求的特征,提出知识的表达方法。根据劣构问题求解的知识需求及分众分类的特点,论证了分众分类在知识分类中的可行性和意义,认为分众分类可以满足劣构问题求解中知识分类的需求,并能促进知识的共享和创新。最后给出基于Tag技术的知识分众分类方法。文章分析了目前企业中进行专家推荐的方法,以及这些方法在实际应用中存在的问题,结合劣构问题求解的知识需求特点,提出基于知识分众分类信息进行专家的推荐策略,认为通过对知识分众分类信息的挖掘,可以有效的解决劣构问题求解中对专家推荐的需求。给出一个无需专家库和专家地图的专家推荐算法,并对算法实现中的一些关键问题做了分析。比较目前常用推荐算法的优缺点及其适用范围,分析知识推荐的特点,研究劣构问题求解中知识的推荐策略。提出基于Tag技术的改进关联规则算法,解决知识学习中的知识推荐问题;提出基于三维协同过滤的知识推荐算法,解决知识应用中的知识推荐问题,通过来自于企业实际的数据集进行实验,结果表明:基于三维协同过滤的知识推荐算法在推荐质量和性能上都优于传统的协同过滤算法。提出面向劣构问题求解的知识管理系统的系统结构,对系统的功能做详细的描述,并分析系统的工作机理,给出基于Agent的系统主动工作模型,设计并开发面向劣构问题求解的知识管理系统的原型。更多还原

【Abstract】 There are many ill-defined and unstructured problems existing in enterprises’decision making, which are named the ill-structured problem. The solving of this kind of problems is important to the survival and development of the enterprises. However, the components and solution space of the problems are unknown which results in several ways or no certain way at all to solve the problems, therefore the ultimate solving method is unforeseeable and emanative and this is complicated for decision makers’knowledge. People usually use brain-storming to solve the ill-structured problem, but the information and knowledge used in the solving process as well as the consequent new knowledge are not summarized and managed, i.e. the knowledge disperse in the individual’s brain. Consequently, how to manage the knowledge in the ill-structured problem solving process efficiently and support the communion of the knowledge to assist the solving are critical to the enterprises. The dissertation studies the ill-structured problem as well as the knowledge management in the solving process, and the main points are as follows:First, the dissertation defines the ill-structured problem including its concept, range, characteristic and taxonomy. Basing on the analysis of activity in solving process, a solving model toward the ill-structured problem is proposed and the general solving process is formalized. The given example in specific enterprise suggests that the model can not only reduce the complexity of ill-structured problem solving but also standardize the solving process.Second, the dissertation analyzes the characteristics and expression of the knowledge demand, proposes a uniform expression of the knowledge and investigates the acquirements of the dominant. According to knowledge demand in solving process and the folksonomy characteristics, it is suggested that the folksonomy is feasible to classify the knowledge. The folksonomy can satisfy the demand of knowledge classification in ill-structured problem solving as well as promote the share and innovation of knowledge. A Tag folksonomy technology is represented in detail.Third, the methods for expert recommendation are discussed. Considering the problems in application and the demand characteristic of solving, an expert recommendation strategic basing on knowledge folksonomy is proposed. The data mining on knowledge folksonomy information can solve the expert recommendation in ill-structured problem solving process. An expert recommendation arithmetic extricating from expert warehouse and expert map is presented and the key points in appliance is also analyzed.Forth, the advantages and limitations as well as applicability of the common recommendation arithmetic are discussed. A knowledge recommendation strategic used in ill-structured problem solving process is proposed to meet the demand of knowledge recommendation. An improved association rules arithmetic basing on tag technology is proposed to solve the knowledge recommendation problem in knowledge study. The 3-D collaborative filtering is extended to recommend knowledge and the experiment using the real data in enterprises is carried out. The results show that the quality and performance of the 3-D arithmetic is better than the traditional arithmetic’s.Fifth, the system structure toward ill-structured problem solving is represented. Further, the system’s functions and mechanism are described in detail. An active working model basing on agent is proposed and the prototype of knowledge management system toward ill-structured problem solving is also designed and developed.更多还原