【作者】 刘东升；

【导师】 琚春华；

【作者基本信息】 浙江工商大学 ， 企业管理， 2008， 博士

【摘要】 随着信息技术的发展,经济全球化进程的加快,以沃尔玛(Wal-Mart)、家乐福(Carrefour)、麦德龙(Metro)等为代表的国际大型连锁零售企业通过20世纪80年代的观望和探测,90年代的潜心“修炼”,已“水土渐服”,已潮水般抢滩中国零售市场。面对拥有雄厚资本与先进管理理念的大型外资连锁零售企业,我国连锁零售企业必须转变管理理念,实施以4C(Customer、Cost、Convenient、Communication)为中心的现代企业管理模式,从分布数据中得到有用的信息、获取分析决策模式和知识,支持连锁商业企业经营决策,将零售业的“商品经营”演绎成“信息经营”,才能提高我国零售企业自身竞争力和发展能力。其中,连锁零售企业客户关系管理CRM系统应用正是实现这一现代管理理念的基础关键之一,但是我国零售业能深层次实施CRM的企业并不多,因此面向连锁零售企业客户关系管理R-CRM(Retail Customer Relationship Management)模型与决策机制的研究具有重大的现实意义和广阔的发展前景。本文在国内外研究基础上,以连锁零售企业商品驱动、供需联动为主线,全面分析并建立了面向连锁零售企业的客户关系管理模型,并就此模型的三个方面进行了深入分析和探讨。本文主要研究内容包括:第一,分析了面向连锁零售企业的客户关系管理管理结构,阐述了连锁零售企业客户关系管理R-CRM的客户消费分类分析、供应商评价分析、企业经营决策机制三个问题的解决思路。第二,提出了面向连锁零售企业的客户关系管理R-CRM框架,对该R-CRM框架的业务流程、三维结构体系、分析指标与方法、功能模块等进行了深入研究,并以此为基础构建了基于商品驱动、供需联动的R-CRM模型,实现对连锁零售企业的消费者、业务伙伴供应商、内部客户等的全方位管理管理。第三,针对连锁零售企业客户消费特点和连锁零售企业商品种类繁多的特性,提出了基于支持向量机(Support Vector Machine)的连锁零售企业客户分类模型:R-DCSS模型,采用云状处理过程的映射机理和非确定性推理,结合SVMDT中的概率分布函数,应用SMO算法对连锁零售企业客户进行分类。最后对该模型在连锁零售企业客户分类中进行了验证,结果表明该模型的分类精度优良。第四,针对连锁零售企业的商品驱动、供需联动的特点与要求,提出了基于BP神经网络的连锁零售企业供应商评价模型,该模型采用主成分法从大量评价因素中筛选出供应商评价的主要因素,即在保留评价信息的前提下对数据进行有效降维,并通过BP神经网络的自学习功能计算出供应商评价的定量评价。通过试验表明该模型具有良好的适用性、准确性,是连锁零售企业供应商评价的一种有效方法。第五,结合连锁零售企业数据分布、异构等的特点,提出了基于贝叶斯网络的连锁零售企业分布式决策模型:R-BNs模型,该模型以贝叶斯网络相关性学习理论为方法,以Bee-gent系统为框架,从分布连锁企业数据库中训练得到迭代型贝叶斯网络,实现连锁零售企业高效、精确的分布式迭代决策。最后对该算法进行了验证,结果证明该算法具有精确、高效并节省空间。最后,设计实现了一个支持连锁零售企业的客户关系管理原型系统:HZ-RCRM。该原型系统结合银泰百货集团的实际需求,将本论文中的改进算法和模型应用到系统中,实现了基于商品驱动的对供应商进行评优选择、客户消费行为分析分类、连锁零售企业的分布式决策等功能。同时,系统对分布式数据挖掘技术在连锁零售企业客户关系管理中的应用也做了有益的探索,为提高连锁零售企业的经营分析、决策支持和商品管理等水平都提供了有力支持。更多还原

【Abstract】 With the development of information technology and the economic globalization, the giant retailers, like Wal-Mart, Carrefour and Metro, have entered and shared the retail market of China. For coping with the challenge of international retailers which hold powerful capital and advanced management ideas, Chinese chain retailers must change old management ideas to actualize the 4C-centered (Customer, Cost, Convenient, Communication) mode of management and transfer "commodity business" to "information business" by collecting and picking up useful data from distributive environment to support decision making. By this way, the abilities of competition and self-grow can be improved for Chinese chain retailers. In the process of retailing management, CRM system is the key to implement the advanced management ideas. The investigation shows that Chinese retailers rarely can implement CRM plan deeply. So, the study on the CRM model for Chain retailers, called R-CRM (Retail Customer Relationship Management) in this paper, is necessary and significant for research and application. Based on the existing research, this paper builds the CRM model for chain retailers in which the linkage between the supplier and the customer is the main idea and the operations in the model are driven by commodity’s. The detail contents in this paper describes as follows:1. The subject of CRM for chain retailers is analyzed and the solutions to customer classification, suppliers evaluation and decision-making in the process ofbusiness management are described.2. The framework of R-CRM is proposed firstly. The business workflows, three-dimensional structure, the index system of analysis and the function modules are analyzed deeply for building the three-dimensional model of R-CRM, in which the linkage between the supplier and the customer is the main idea and the operations in the model are driven by commoditys. By this way, the customer, the business parterner and interior clients can be management all-round and efficiently.3. Aiming at the characteristics of retail customers buying and the variety of retail commoditys, the customer classification model based on SVM is proposed, called R-DCSS. In the model, the cloud-form mapping mechanism, the nondeterministic inference, the probability distribution function and the algorithm of SMO are used to classify the buying behavior of customers. The experimentation shows that the classification precision of the model is excellent.4. Based on BP neural network, the supplier evaluation model is proposed considering the demands of commodity-driven and the linkage management between suppliers and customers. Using the principal component method to choice main factors of evaluating suppliers, the number of dimensions can be cut down on the preconditions of saving complete evaluation imformation. Then the quantitative evaluation to suppliers can be figured out by self-learning in BP neural network. The experiment shows that the model can work with good adaptability, effectiveness and accuracy.5. Considering the distribution and heterogeneity of retailing data, the distributive decision-making model based on Byasian network: R-BNs is proposed. In the model, the correlation learning theory is applied and the Bee-gent system is used to build the framework of the model. Using the data in the distributive database, the iterative byasian network can be obtained by training to support the business decision-making. The experiment shows that the algorithm holds the characteristics of precision, efficiency and space-saving.6. For application purposes, a CRM propotype system for Hangzhou Department Store is designed and applied considering the management demands of the retailer, called HZ-RCRM. Benefiting from the algorithm and model in R-CRM, the Chinese retail enterprise can choose the optimum supplier driven by commodities, classify customers considering their buying behavior, and achieve distributive decision-making support. Furthermore, the efficiency of management and the accuracy of decision support can be improved.更多还原