【作者】 区玉明；

【导师】 张师超； 严小卫；

【作者基本信息】 广西师范大学 ， 计算机软件与理论， 2004， 硕士

【摘要】 随着Internet在全球迅速扩张并逐渐渗透到了社会各个领域，以Internet为平台的电子商务得到了蓬勃的发展。在电子商务中，客户保持是一个重要的主题，它与企业的生存和发展是息息相关的。以客户为中心的现代客户关系管理认为，客户是企业最重要的资源，也是企业赖以生存和发展的利润源泉，企业追求的真正目标是实现客户的长期满意，而不是一次性的交易。只有提高企业的客户保持能力，与客户建立起长期、稳定的关系，才能增强企业的竞争能力和持续发展能力，从而使企业在激烈的商业竞争中立于不败之地。企业要想获得客户的长期满意，提高客户的忠诚度，实现客户保持，必须具备从深层次上分析和理解客户的行为、偏好以及需要的能力。而Web挖掘技术恰好是一个能够为企业提供这种能力的有力工具。Web挖掘是从数据挖掘发展而来的，它能够从大量的Web数据中发现隐含的、有价值的和可应用的模式和知识来为企业决策（比如：客户吸引策略、客户保持策略）提供帮助。本文利用Web挖掘技术针对客户保持中的下列两个重要问题进行了研究。（1）对网站感兴趣用户的识别问题。对于一个电子商务网站来说，每天的访问量都是巨大的，并且访问者的目的也是不尽相同的。如果我们把这些形形色色的访问者不加以区别地统一看待的话，那么我们所分析出来的购买者模式必将是非常不精确的，对于企业的客户保持策略的制定不但没有多大的实际帮助，甚至还会误导企业的决策。为此，我们提出了一个对网站感兴趣用户的识别技术，把那些对网站（商品）感兴趣的用户识别出来，以提高下一步分析工作的效率、精度，为客户保持决策提供具有可用性的分析结果。其基本思想是，由于在Web日志中详细记录了访问者对网站访问的详细信息，我们基于Web日志，利用分类的思想，根据他们的访问特点，选择了9个属性通过决策树学习构造了一个分类器，对访问网站的所有用户进行分类，把他们分成三类：对网站感兴趣用户、对网站不感兴趣用户和网络机器人，从而达到识别出对网站感兴趣的用户的目的。（2）网站质量评价问题。随着电子商务的应用越来越普及，企业的网站已不再只单纯作为企业对外的一个门户，它更多的成为了企业进行商业活动的经营场所。一个质量好的网<WP=6>站，将会大大方便客户对网站信息的获取和商品的购买，从而使客户对网站（企业）产生好感，提高客户的忠诚度，达到客户保持的目的。因此，企业网站质量的好坏是至关重要的，它已经成为了企业在日益激烈的商业竞争中取得胜利的一个重要砝码。但如何才能得知自己的网站质量如何、通过哪方面来提高网站的质量以及与竞争对手的网站相比又如何呢？为了解决这些问题，我们提出了一个网站质量评价系统，其基本思想是，先定义了一些属性，接着通过Web爬行技术和Web网页信息抽取技术获得各属性的取值，然后利用映射函数把各属性值映射到[0，1]区间上达到对各属性单独评价的目的，最后对这些属性的评价值进行整合而最终完成对网站质量的评价。我们根据我们提出来的解决方法，实现了原型系统，并用真实数据进行了实验。实验结果表明，我们的方法是有效、可行的。更多还原

【Abstract】 In E-Commerce, customer retention is one of the most important topics, which is vitally important to enterprise’s development and success. From the point of view of CRM (Customer Relationship Management), customer is the most important enterprise’s resource. Enterprise must build long-term relationship with its customers to enhance its competition ability and realize its sustained development.To gain the customer’s long-term satisfaction, increase customer loyalty and retain customer effectively, enterprise must have the ability to understand deeply the customer’s behavior, preferences and requirement. Web mining, which develops from Data mining, is just one of the most prevailing techniques that can provide this ability for enterprise. It can discover valuable and actionable information from a huge amount of Web data to assist in enterprise’s decision-making, such as acquiring new customer and retaining old customer.In this paper, we study the following two issues of customer retention by using Web mining technique.1. Identifying visitors who have purchase interest. There is a large number of visitors to an E-Commerce website everyday. But their aims are quite different. If we don’t distinguish them when we analyze their visit patterns, we will get the imprecise results that will not help enterprise for decision-making. To solve this problem, we present an approach to identify visitors who have purchase interest. Our basic idea is that we make use of the information contained in Web-Logs and a classifier consisting of 9 attributes to classify the visitor at a certain website into three different groups, that is, visitors who have purchase interest, those who have no purchase interest, and the network robots.2. Website quality evaluation. With the population of E-Commerce, website is used not only initially as a “window” of an enterprise that customer interact with, but also gradually as an important environment for trading merchandise. A website with good quality is convenient for <WP=8>customer to obtain information and purchase merchandise, and consequently it helps enterprise to increase customer loyalty and achieve customer retention. So the quality of website is of importance. However, how does an enterprise know the quality of its website and the situation compared with its competitors? We present a website quality evaluation method to solve this problem. The method basically consists of the following steps: (a) Specifying object to be evaluated; (b) Selecting some attributes for evaluating; (c) Getting the value of all attributes; (d) Defining evaluation functions and evaluating the attributes; and (e) Yielding the global evaluation by aggregating all the evaluations of attribute.We build a prototype according to our presented methods. The experimental results show that our methods are effective and feasible.更多还原