【作者】 刘平峰；

【导师】 聂规划；

【作者基本信息】 武汉理工大学 ， 管理科学与工程， 2006， 博士

【摘要】 现有电子商务推荐系统不能满足网络条件下大规模电子商务推荐的需求，知识网格技术、语义本体论和电子商务推荐技术的结合，可以满足网格条件下商品知识、用户需求知识和推荐知识的有效获取、聚合和智能推荐协同工作的要求。通过对基于知识网格的电子商务智能推荐的理论与方法进行研究，为大规模、高质量和强实时性要求的分布式电子商务智能推荐系统的研发提供理论方法基础。本文将基于知识网格的电子商务智能推荐理论方法研究分为三级结构：基于知识的电子商务智能推荐系统、电子商务智能推荐知识网格、以及电子商务智能推荐知识网格服务社区，分别研究了基于知识的电子商务智能推荐原理方法；研究设计了电子商务智能推荐知识网格模型，研究设计了电子商务智能推荐知识网格服务社区的结构、生成和组织机制以及自组织优化算法。论文主要研究内容如下：相关理论分析与评述。主要阐述了语义Web的基本理论，网格技术的最新发展，国内外主要流派关于语义网格和知识网格的基本思想、理论、技术和方法。分析了电子商务推荐系统的概念，基于内容过滤和基于协同过滤的推荐技术，以及推荐系统中涉及的数据挖掘技术。基于知识的电子商务智能推荐(KBIECR)原理方法研究。研究设计了KBIECR系统的逻辑框架及其工作原理；结合电子商务推荐领域本体RecommendOnto的构建，研究了推荐系统中用户知识、商品知识、规则知识和案例知识统一表示的方法；在讨论KBIECR知识获取一般原理方法的基础上，重点探讨了商品知识和用户知识的获取方法；研究了KBIECR本体推理、规则推理和案例推理三种推理机制，设计了用户需求语义相似度算法，探讨了推荐结果基于语义相似度和商品特征熵的产生方法；最后采用ART神经网络实现推荐系统中推荐策略的自适应性。电子商务智能推荐知识网格(IECRKG)模型研究。研究设计了电子商务智能推荐知识网格的层次模型和基于服务的架构，探讨了IECRKG资源空间模型和资源统一表示的方法，设计了IECRKG资源操作语言，对实现语义交互环境的关键技术，特别是基于语义的服务匹配和基于知识的工作流计划技术进行了深入的研究，最后对IECRKG知识搜索服务、知识发现服务和知识聚合服务的实现进行了探讨。电子商务智能推荐知识网格服务社区研究。研究了IECRKG服务社区的基本原理、结构、生成和组织机制，设计了服务社区基于语义的关键路由算法以及服务社区自组织优化的算法。更多还原

【Abstract】 Current recommender systems cannot meet the demand of large-scale electroniccommerce in the distributed environment. The knowledge engineering, semantic webtechnologies and electronic recommendation technologies together can be exploited torealize the effective acquisition and fusion of the knowledge of products, userrequirements and recommendation function, and can further realize the synergizedcollaboration of constituents of the recommender systems in the grid environment.This dissertation researches on the basic theory and methodology of intelligentelectronic commerce recommendation based on the knowledge grid which we hopewill provide the fundamental support in theory and methodology for developinghighly responsive and effective recommender systems for large-scale electroniccommerce in distributed environments.The research on the theory and methodology is subdivided into three levelswhich are correlated with each other closely to support distributed recommendationfor large-scale electronic commerce. The first is the theory and methodology ofknowledge-driven intelligent electronic commerce recommender systems. The secondis the theory and methodology of intelligent electronic commerce recommendationknowledge grid (IECRKG). The third is the theory and methodology of the IECRKGservice communities.In chapter 2 the relative theories are introduced which include the basic theory ofthe semantic web, current developments in grid technologies, the theories,methodologies and technologies for semantic grid and knowledge grid, the conceptand main technologies of electronic commerce recommender systems and theconcerned data mining technologies.In chapter 3 the theory and methodology of knowledge-based intelligentelectronic commerce recommender systems are proposed. Firstly how to represent theknowledge in recommender systems uniformly is explored based on the constructionof RecommendOnto, the ontology for the domain of electronic commercerecommendation. After the general analysis of the methods of acquiring knowledge inrecommender systems, the methods of acquiring knowledge of products and userrequirements are elaborated. The mechanisms of ontology reasoning, rule reasoningand case based reasoning are discussed. The method of generating recommendationresults is also discussed by adopting the proposed algorithm of semantic similarityand entropy of commodities’ features. Finally the self-adaptive algorithm of choosingrecommendation techniques in recommender systems is analyzed. In chapter 4 the hierarchical model and service oriented architecture of IECRKGare designed. After that the resource space model, the uniform representation ofresource, and the IECRKG resource operation language are proposed. Then the keytechnologies concerned with the semantic interaction, especially the semantics basedservice matching and knowledge based workflow, are explored. Finally how toimplement knowledge search service, knowledge discovery service and knowledgefusion service is discussed.In chapter 5 the problems of how to organize and optimize the communities onIECRKG are tackled. The main work includes: the principle and structure of thecommunities, the organizing and formalizing mechanism of the communities, the keyrouting algorithms, and the algorithms of self organized optimization.更多还原