【作者】 金绳轩；

【导师】 邵世煌；

【作者基本信息】 东华大学 ， 控制理论与控制工程， 2007， 硕士

【摘要】 纺织服装快速反应系统是近十年来兴起的一种产业链上下有机衔接、信息资源共享、快速响应市场需求的企业联合运作模式。纺织行业面临着：①市场变化快，客户需求日显个性化、多样化；②国际化、一体化进程加快；③信息网络日趋完善，信息交互快。纺织快速反应的关键恰恰是一个“快”字，即不管客户何时何地提出何种要求，也无论对方所需的数量和类型，都能毫不延误地组织生产，最大限度地满足对方需求。纺织快速反应的另一个关键就是一个“应”字，而“应”的对象就是“信息”。在快速反应系统中，变化的起源在于信息的变化。要想准确、及时地把握市场脉搏，必须建立一个快速反应信息系统。如何实施快速反应、在已有纺织信息系统的基础上建立快速反应信息系统，这对中国纺织服装企业来说是一个亟待解决的问题。本文就是围绕如何借助信息技术建立服装企业的快速反应机制开展研究的，对纺织服装企业快速反应信息系统的建立和实施过程中的若干关键技术进行了分析研究，并构建了纺织服装快速反应信息系统框架。本文的主要研究内容与成果有：1．在纺织快速反应信息系统的平台实施技术研究中，提出了一套基于WEB的面向纺织服装企业信息集成解决方案，该方案能针对纺织服装相关主题进行信息的自动提取与分类。在设计信息抽取规则时，提出了一种改进的纺织服装企业抽取规则，该规则将“基于文档结构”与“基于特征模式匹配”二者结合，能较好的提高信息抽取的准确度与可移植性。在实现文本分类时，给出了一种改进的KNN快速文本分类方法并构造了一种基于类中心的多维索引结构，该方法可以在海量数据集中的情况下进行快速有效的分类。2．在纺织快速反应信息系统的个性化信息服务技术研究中，着重分析了基于数据挖掘的个性化服务，并以服装品牌忠诚度为例，提出了ID3分类算法和K-means聚类算法相结合的个性化数据挖掘算法：采用ID3算法来构建决策树，改进了“属性信息增益”的计算处理方法，提高了算法的效率设计；采用K-means算法来实现针对服装品牌忠诚度的客户聚类。最后给出了算法实现，包括挖掘算法的发布、发现、调用等。3．在纺织快速反应信息系统的在线品牌诊断技术研究中，首先建立了“品牌特征信息-品牌诊断信息-品牌价值”的品牌诊断层次关系，推导出一个关于诊断对象、特征信息、诊断信息、映射关系及案例集的五元组诊断模型：DC=f(DP)=f(O，CI，DI，T，E)，提出了基于双层嵌套推理框架的层次聚类推理方法来进行诊断推理：在案例层，定义广义相似度公式来进行品牌案例匹配；在规则层，采用模糊算法来实现规则推理，在设计诊断模糊矩阵R时，采用神经网络算法自动修改权值结合人工检验的方法以确保其准确性。更多还原

【Abstract】 Quick response system is an enterprise’s unite operation model rised in recently ten years, which links up the industry chains, commonly shares information resources and rapidly responds market demand. A textile enterprise is facing a market of individuation and diversification; the course of internationalization and integration quickens; also with the consummation of the network,the information communications speed up. A textile&costume enterprise ought to cope with this challenge. However, the key word of quick response for textile is just "quick" .It desiderates resolving how to achive quick response, make produce in time basing on consumers’ need, and quickly occupy the market to win in competitions.On the other hand, another key word is "response" ,whose object is "information".In a word, the origin of change rests with the variety of information.We should set up a quick response information system in order to grasp the market’s opportunities.In this article, we have analyzed how to build up the quick-resonse strategy by means of information technologies and the key implementing technologies in the process of forming the quick-response- information-system, and then construct the sketch of information quick-response-systems for textile and costume.The main content and innovation of this article composes of three parts:1. Put forward an information intergration solvement based on textile& costume enterprises,which provides an approach for automatic information extraction and classification based on textile subjects.Put forward an improved extraction rules,which combine the information extraction methods based on "document structure" with that based on "characteristic pattern matching ".Also,in this article,we provide an improved method for rapid -text-categorization of KNN,which can provide a quick search among the mass data set.2.In the research of technologies for individuation services,the principle of individuation services based on data-mining is analyzed.Then an individuation data-mining arithmetic, which combines ID3 classification arithmetic and K-means clustering arithmetic,is analyzed.At last, the realization of the arithmetic is provided.3. A diagnosis structure of costume brand is established by discussing the modeling process in detail. The diagnosis model is given as followed: DC = f (DP) = f(O,CI,DI,T, E).On the basis of that, a hierarchical clustering reasoning method ,combining RBR and CBR, is introduced sequently. Meanwhile,the reasoning mechanism is designed according to the procedures of brand diagnosis and its corn arithmetic.更多还原