【作者】 陈敏；

【导师】 刘晓强；

【作者基本信息】 东华大学 ， 计算机应用技术， 2008， 硕士

【摘要】 现代企业80%的利润是来自20%的重要客户,而其余80%中的大部分客户对企业是微利的,甚至是无利可图的。因此如何建立高效的客户知识管理,实施合理的客户分析,有效地获得企业的关键客户,从而对企业客户实施差异化管理,已经是各企业急需解决的课题。本文提出了一种基于客户知识进行客户分析的CRM系统的实现方案,通过获取客户知识、分析客户知识,利用数据挖掘的相关方法、流程建立基于客户知识的客户分析模型,对客户进行分类,为企业营销活动提供支持。本文根据CRM中数据挖掘的流程,对客户分类数据挖掘过程进行实例化:确定模型的问题域——实现客户细分;根据实现客户细分的主题,展开数据准备工作;建立数据挖掘算法;实施数据挖掘,得到分类结果。对于数据挖掘算法,由于客户分析中客户类别预先未知的特点,我们选用经典聚类算法——基于距离的K-Means聚类作为分类算法。同时引入了模拟退火策略的思想,改进K-Means聚类算法,从而解决K-Means在处理孤立点数据容易造成聚类结果局部最优化的问题。在进行聚类分类的过程中,由于客户知识分布零散、粒度太小的特点,无法直接将其作为聚类距离进行分类。本文通过建立一个客户综合评估模型,从零散的客户知识中抽取可供评估的知识,并将其降维、集成为一个可直接用于评估的综合指标－－客户综合评估值,作为聚类的分类距离,为数据挖掘提供数据依据。在建立客户综合评估模型过程中,本文采用了无量纲化的方法对客户知识的多个指标进行规范化处理,并采用层次分析的方法来综合这些指标,实现降维处理。论文最后设计并实现了基于客户知识的CRM系统,建立了一个完整的客户知识获取、客户知识存储、客户知识应用并实现客户知识创新的系统。通过客户交互子系统为企业提供多种途径来获取客户知识;通过业务应用子系统和客户分析子系统来应用客户知识;通过客户知识库管理子系统来共享、创新客户知识,子系统不仅为企业用户提供了合理客户分析的评估模型模版、算法模版,还提供用户定制客户分析评估模型库和算法的功能。更多还原

【Abstract】 Nowadays, modern enterprises make 80% of their profits on their 20% key customers while the remaining 80% of the customers have made little or none profits to the enterprises. In order to obtain the greatest degree of profit, the enterprises may need to implement dynamic and different customer management, and make the survival of the fittest customers. This will not only help enterprises reduce customer costs and enhance customer profitability, but also spend enterprise resources on the 20% key customers, which will give the core interests for the enterprises, to maximize the interests of enterprises. Therefore, to establish a highly efficient customer knowledge management, implement a reasonable customer analysis model and effectively gain the key customers, thus implement different management for the enterprise customer, is the urgent problem for business.This paper brings forward a knowledge-based CRM system, which can provide analysis support for the enterprises. The system can access, analyze the customer knowledge and build a knowledge-based customer analysis model by using data mining methods, so as to implement customer classification and provide reliable basis for enterprise marketing activity. According to the process of data mining in CRM, we give the example of implementing customer classification as follow: Firstly, identify the problem domain - the customer segmentation; Secondly, prepare the data; Thirdly, establish the data mining algorithms; Fourthly, implement data mining.During the process of establishing data mining algorithms, as the types of customers are unknown in advance, we use a classic clustering algorithm -- distance-based K-Means clustering algorithm as a classification algorithm. At the same time the author introduces simulated annealing strategy thinking on the traditional K-Means clustering algorithm, so as to improve the traditional algorithm and solve the partial optimization problem caused by isolated data points during the clustering process.During the process of preparing the data, the customer knowledge can not be used directly as the classification distance for clustering, as they are scattered, small in particle size. This paper builds an integrated assessment model to draw indicators out from the fragmented customer knowledge and integrate these indicators into a synthetical indicator by using dimension reduction method. The synthetical indicator is named as customer comprehensive assessment value, which can be used as the classification distance in clustering. During this process, the author uses the dimensionless method to standardize the indicators and then uses AHP method to integrate the indicators.At the end, this paper designs and establishes a knowledge-based CRM system. With the aim of achieving the goal of knowledge innovation, the system implements the functions as follow:Through the enterprise portals internet to provide multiple channels for enterprise to obtain the customer knowledge.Through the business application subsystem and customer analysis subsystem to apply customer knowledge.Through the customer knowledge base management subsystem to share, innovation customer knowledge.The customer knowledge base management subsystem not only provides the model and the algorithm template for customer analysis, but also provides enterprise functionalities of customizing customer analysis model and algorithm.更多还原