【作者】 陈获帆；

【导师】 赵学锋；

【作者基本信息】 华中科技大学 ， 管理科学与工程， 2009， 硕士

【摘要】 随着零售电子商务的快速发展,在线信誉管理系统的研究越来越受到学者们的重视,目前一些简单的在线信誉管理系统已成功地运用于众多C2C电子商务网站以及一些B2C购物代理网站,但是目前的在线信誉管理系统的维度设计还不够完善,针对上述存在的问题,本次研究将从客户评论的角度,采用文本挖掘的方法来研究B2C在线零售商的信誉维度,从而对目前主流的B2C电子商务网站的信誉维度进行优化。本次研究利用文本聚类技术对客户文字评论进行处理与研究,主要可分为两大部分,第一部分为文本转换,它可以分为三个步骤:(1)文本集合的生成;(2)特征项集合的生成;(3)VSM数值矩阵的生成和优化。通过这三步,我们可以将大量复杂的文档转换成可以被计算机直接处理的数值矩阵,为聚类分析奠定了基础,其中,第二步和第三步是我们的研究重点,包括特征项选择算法,权重函数的确定等方面的研究。第二部分为聚类分析与应用,这一阶段由两步组成:(1)将生成的数据矩阵进行聚类分析,得出聚类结果。(2)对聚类结果进行评价检验,并应用到相关领域。在聚类过程中,我们将采用层次聚类和k-means聚类相结合的方式,用层次聚类算法作为主要的聚类手段,而用k-means聚类算法进行迭代检验。在得出聚类结果之后,我们将进行知识提取,并应用到相关领域。通过本次研究我们可以发现聚类分析在电子商务中的应用是可行的,并且具有很重要的意义。这是一种新的信誉维度确立方法,具有一定的科学性和合理性。除了确立在线零售商的信誉维度,我们在聚类过程中还可以发现不同客户群体和不同零售商群体的典型特征,从而制定出差别化的客户服务方案等。随着统计技术与计算机技术、人工智能技术的紧密结合,新的面向具体应用领域的、具有弹性的聚类分析技术和应用软件将会层出不穷,其解决问题的广度和深度将会得到更大的提高。更多还原

【Abstract】 With the rapid development of retail e-commerce, online reputation management system is attracting a lot of attention of the scholars. At present, some simple online reputation management system has been applied to many C2C and B2C e-commerce sites successfully. However, there are still some shortcomings in online reputation management systems’reputation dimension design .To solve the above problems, we should research the reputation assessment dimension of B2C online retailers with the method of text mining in this paper, to optimize the reputation dimension of the mainstream B2C e-commerce site ,which from the view of customers’perspective.This paper process and research the customers’perspective with text clustering technology can be divided into two major sections. The first section is the text-number conversion, it can be divided into three steps: (1) The generation of text collection; (2) The generation of Characteristics collection; (3) The generation and optimization of VSM numerical matrix. Through the three steps, a large number of complex documents can be converted to numerical matrix which can be processed by the computer directly, and it would lay the foundation for cluster analysis, of which, the second and third step is the focus of our research, including the characteristics collection algorithm and the VSM weighting function study. The second section is cluster analysis and application, it can be divided into two steps: (1) Data matrix processing with clustering analysis. (2) Clustering analysis results text and related fields application. In the clustering process, we will combine the hierarchical clustering and the k-means clustering, with hierarchical clustering algorithm as the main means of clustering, and using k-means clustering algorithm for iterative testing., we will extract knowledge and applied to related fields after the findings of the clustering results.We can found that the application of text clustering analysis in e-commerce is feasible through this study, and it is of great importance to us. This is a new method to establish the reputation dimensions, it is of science and rationality. In addition to the established of the online customers’reputation dimensions, we can also found the features of different customer groups and different online retailer groups through text clustering, and then develop a differentiated customer service programs. With the closely combination of statistical technology and computer technology and artificial intelligence technology, the new flexible cluster analysis techniques which for specific application and software will be endless, the solution to the problem will be greater improved on its span and depth.更多还原