【作者】 于小兵；

【导师】 郭顺生；

【作者基本信息】 武汉理工大学 ， 工业工程， 2010， 博士

【摘要】 网络技术与信息技术的飞速发展以及3G(Third Generation)的推广普及使得电子商务、网上贸易已经深入到企业管理、社会生活的各个方面。电子商务凭借着Internet平台,给人们带来了极大的便利和取之不尽、用之不竭的资源。但是,资源的丰富性同时也给用户和电子商务企业带来了很多难以解决的问题：如何为电子商务客户提供个性化服务、如何洞察客户的消费行为、如何提高广大中小电子商务企业的竞争力。这些都是智能电子商务IEC (Intelligent E-Commerce)必须面对和解决的问题。本文通过数据仓库和移动Agent技术,建立了IEC的体系结构：通过数据仓库,进行挖掘分析,形成预测和知识,以便于企业业务决策；借助于移动Agent,实现与加盟网站的信息交互,以提高网站的信息丰富度和竞争力。具体工作如下：(1)提出了RMFRM (Refining Maximal Forward Reference Mode)算法用于获取用户事务序列。结合网站的有效页面编码,将用户事务序列转化为二进制向量。对传统的蚁群聚类算法的六条规则进行改进,提出了改进蚁群聚类算法,用于二进制向量的聚类,从而实现用户的聚类,为个性化服务和推荐做好了准备。(2)预测了电子商务交易量。鉴于电子商务的交易量对电子商务的运作有着决定性的影响,给出了预测前的数据完整性检查算法。通过极端学习算法预测了电子商务的交易量,取得了较高的预测精度。(3)提出了集成有序加权平均OWA(Ordered Weighted Averaging)和粗糙集求解电子商务客户流失权重。探讨了电子商务客户生命周期及其流失预测模型。针对客户流失给电子商务企业带来的严重损失,从专家和用户二个角度出发,通过OWA和粗糙集,全面细致地讨论了流失权重的计算过程。通过案例,指出退货及售后服务是导致客户流失的主要原因,给出了相应的补救措施。(4)构建了基于移动Agent的电子商务联盟体系结构。给出了基于模糊支持向量机SVM (Support Vector Machine)的加盟算法。为了提高电子商务联盟的运作水平,提出了基于模糊理想点法TOPSIS (Technique for Order Preferenceby Similarity to Ideal Solution)算法的电子商务竞争力评估算法。(5)开发了基于多层的IEC系统。应用于国外某实际运营的电子商务系统中,取得了良好的效果。更多还原

【Abstract】 Network and information technology are developed very rapidly. Besides, the third generation is promoted and very popular. All these efforts have made e-commerce and internet trade into every aspect of enterprise management and society. E-commerce offers us a lot of inexhaustible resources based on internet platform. However, too many resources also cause some problems to customers and e-commerce enterprises. For instance, how to offer personalized services, how to identify customer consumption behavior and how to improve competition ability for many small and media e-commerce enterprises. IEC has to face and solve these problems.IEC architecture is established based on database warehouse and mobile agent. On the one hand, database warehouse is built up to execute data mining and form prediction report, which can help enterprise make decision. On the other hand, e-commerce websites can exchange information with other websites by mobile agent so that the website can enhance information abundance and improve completion ability. The innovative results are achieved as following:(1) Refining maximal forward reference mode is proposed and used to acquire user access transaction sequences. These sequences are converted to binary vectors combined with the effective page number. Six rules of ant cluster algorithm are extended to form a new algorithm. These vectors are clustered by the new algorithm, which can realize users cluster and make preparation for personalized services and recommendation.(2) E-commerce transaction volume is predicted as it is very crucial to e-commerce. Data integrity algorithm is presented before forecasting. Extreme learning machine algorithm is proposed to predict the transaction volume and result indicates that the approach has much higher precise.(3) E-commerce customer churn weight is solved by OWA and rough sets. Customer lift cycle and churn prediction model are discussed. As customers losing can make enterprise great losses, OWA and rough sets approaches are presented from the point views of experts and users. The calculation process of churn weight is demonstrated very clearly. The case analysis shows that return and sales services are main reasons for customer churn. Based on results, some measures are proposed.(4) E-commerce alliance architecture is proposed based on mobile agent. Fuzzy SVM is presented to decide whether a website can join the alliance or not. In order to improve the operation of alliance, fuzzy TOPSIS is used to rank the competition of e-commerce.(5) N layers IEC software is developed. It is applied to a commercial website. The result demonstrates the reasonableness of the prototype.更多还原