综合决策支持系统中计算智能和知识获取技术的研究与应用

【作者】 郑骁庆；

【导师】 邹平；

【作者基本信息】 昆明理工大学 ， 管理科学与工程， 2003， 硕士

【摘要】 随着经济的发展，商业竞争日益激烈，企业环境千变万化。企业能否快速和正确地对其外界或内部的各种变化做出积极的反应并制定正确的决策，是企业争取时间，迅速行动，从而最终取得市场竞争优势的关键。在信息技术飞速发展，信息数据迅猛增长的今天，企业快速有效的分析和决策越来越依赖于其信息系统的设计与开发。其中，决策支持系统(DSS)是信息系统领域的一个重要分支，是结合计算机科学、运筹学、管理科学等多学科的新型学科。DSS经过设计、构造、用于辅助决策过程。已经成功地用于解决各种各样的问题，在企业决策、城市交通管理、保健及医疗诊断、环境规划、农林业等应用领域发挥重要的作用。人工智能，数据挖掘和数据仓库技术已经对决策支持系统的研究和发展产生了重要的影响。此外，模糊集、粗糙集、人工神经网络和遗传算法等理论与方法也被运用于解决决策有关的问题。各种技术的相互渗透和结合日益明显与复杂，决策支持系统中充分融合各项技术在近年来受到普遍关注和充分重视。 本文撇开众多对知识定义上的争论，从实用性和可操作性的角度对解决决策问题的知识提出了新的认识和分类思想。然后在这种思想上，在一定程度上研究和总结了人工智能、数据挖掘、神经计算、进化计算、运筹学、Rough集和Fuzzy集等理论和方法的研究成果，综合比较了各种技术的本质、特征、异同和优劣。论述了最新科技发展对决策支持系统的影响，并讨论了如何将这些技术有创新性地结合运用于DSS的设计与构造中去，以期对决策支持系统的研究与开发有指导的意义和肩发的作用。 首先，本文在综述与决策支持系统有关的计算智能与知识获取研究动态和进展的基础上，对解决决策问题的知识提出了一些新的认识和分类思想。在此基础上，对各种处理知识典型技术进行重新划分和归纳。进一步明确了处理DSS中知识有关问题的理论与方法的范围和方向。 其次，以计算理论中计算复杂性讨论为基础，分析了决策中所涉及的NP问题，研究了运筹学方法、人工神经网络、遗传算法等方法在解决NP问题时的本质、特点和优缺点。并结合实际中典型的旅行商问题进行实例分析，概括了实际问题求解应如何选择各种计算技术。 第三，在本文对知识认识和分类的基础上，讨论了知识获取中典型的推理和分类的理论与方法。主要分析了数据挖掘、机器学习、粗糙集、神经计算、遗传综合决策支持系统中汁算智能和知识获取技术的研究与应用摘要算法等的技术的实质、优劣和适用范围。提出了新的结合方式和设计了新的决策支持系统结构模型。 第四，结合上述研究的结果，设计了企业决策支持系统中应收账款管理子系统。体现各种技术之间相互融合、取长补短的思想。 最后，综合了上述分析和研究的基础上，对各种理论与方法之间的相互渗透和结合进行了阐述，并展望了人工智能的发展方向。更多还原

【Abstract】 With the development of economy, how to make correct and efficient decision under the frequently changing and intensely competing condition is the key to obtain predominance at the market. As a result of information techniques’ development and progress at very fast speed, making decision in business more and more rely on Information Systems’ design and implementation. Decision Support System (DSS) is an important branch of Information Systems used to support managerial work and decision making. DSS integrates techniques of Compute Science, Operation Research and Management Science and already uses widely at decision making in business, city traffic management, health care and medical diagnosis, environment programming and agriculture management, etc. Recently, Artificial Intelligent, Data Mining and Data Warehouse techniques have made a strong impact on research and resign of DSS; otherwise, Fuzzy Set, Rough Set, Artificial Neural Nets, Genetic Algorism and so forth have been used to solve the decision problem. Due to increasing trend of various subjects’ integration and connection, application of multi-techniques in DSS’ research and design have been roused widely attention.On the basis of discussing different knowledge definition and cognition, this dissertation brings about innovative definition and classification of knowledge for decision making from practical aspect. And then, through comparing methodology and techniques of Artificial Intelligent, Date Mining, Computational Intelligence, Operation Research, Rough Set, and Fuzzy Set, etc. and analyzing their essences, characters, similarities and differences, and advantages and disadvantages, etc. this dissertation puts forward some innovatively combinative methods of vavious methodology and techniques which have certain meaningful direction and enlightened idea in research and design of DSS.Firstly, on the overview of the trend and development in Computational Intelligence and Knowledge Acquisition, this dissertation brings aboutinnovative knowledge definition and its classification. This definition and classification help to specify range of methods and techniques aboutknowledge when building and designing DSS.Secondly, on the basis of discussing Computational Complexity and NP-complete problem, this dissertation studies some methods to deal with NP-complete problem which include Dynamic Programming, Backtracking and Branch-and-bound Algorithms, Local Improvement and Simulated Annealing Algorithms, Hopfield Neural Nets, Genetic Algorithm and etc. Through analyzing various algori thms to solve the Traveling Salesman Problem, how to choice the suitable algorithms when dealing with NP-complete problem has been put forward.Thirdly, on the ground of discussing some representative techniques of reasoning and classification which involve Data Mining, Machine Learning, Rough Set, Fuzzy Set, Neural Computation and Genetic Algorithm, these techniques’ essences, characters, similarities and differences, and advantages and disadvantages ;have been analyzed and innovative structure of DSS has been introduced.Fourthly, receivable maniagement subsystem of Decision Support System has been designed on the bas;is of above research and discussion.Finally, through review:integration of various methodology and techniques, this dissertation has a prospect to Artificial Intellignece in theory frame;更多还原