动态客户关系管理下隐马尔可夫建模及实证

【作者】 严博；

【导师】 邵培基；

【作者基本信息】 电子科技大学 ， 管理科学与工程， 2006， 硕士

【摘要】 有效的客户关系管理是公司保持竞争优势的主要途径。目前公司所关注的客户关系管理的主要内容有:识别、细分、激励、维系和挽留客户。这些内容可以简化为两个问题:谁会买我的东西?将会买多少?为了解决这些问题,众多学者从客户的购买行为出发,计算客户价值。最先的模型单纯计算客户已经完成的购买行为及下一次可能的购买概率。此后有学者提出考虑其他因素的模型,如:企业采取营销措施、客户满意度等。然而随着客户关系管理的广泛应用,公司根据实际提出更多要求。他们不仅希望知道下一期是否降价。更希望知道一年采取几次营销措施,将一年的返款金额定在多少,将使所获客户价值最大。这就需要研究深入到多个时期,解决客户和企业动态决策问题,即动态客户关系管理。目前多期模型的研究中大多采用马尔可夫模型,该模型的特点是很好的解决了客户在多期下购买行为转换的问题。然而该类模型在衡量企业与客户关系时,无法扩展到三个及以上维度。本文引入隐马尔可夫模型,该模型通过观察客户的属性及购买决策序列,对客户与企业的关系状况进行较全面的衡量;可全面分析各类营销活动对客户全生命周期价值(或客户资产净值)的整个影响(即当期的影响和以后带来的影响)。该模型有以下特点:(1)目前动态客户关系管理中,基本以交易额为标准衡量客户与企业的关系;在新模型中可以引入更多指标来衡量,而不受数量限制。(2)对客户与企业的关系本身就是一种分类,故新模型能形成对客户的动态分类。此外,本文对网上零售企业进行了实证分析。通过客户的历史购买数据计算出单个客户状态序列及单个客户的响应序列。为企业提供了一种客户行为的定量研究方法,为该企业开展个性化客户关系管理提供了基础。更多还原

【Abstract】 The effective customer relationship management is the main route that the company keeps the competition advantage. The main jobs of customer relationship management which companies considered include four parts, such as, identify, subdivision, promote, maintain customer. The contents can be simplified to two problem: who will buy our products? How many he/she will buy? To solve the problem, many scholars focus on the customer’behavior for calculating the customer’s value.The early model only calculate the behavior data which customers has, and predict the purchasing probability next term. Base on this model, some scholars consider more factors to calculate the purchasing probability, such as, sales promotion and customer satisfaction. In practice, however, present studies didn’t match the needs which the companies face. The company not only want to know whether cheapen next term, but also want to know how may times he needs to carry out sales promotions. So to solve the problem, we need the dynamic customer relationship management.This dissertation addresses the issue of modeling and understanding the dynamics of customer relationships. The proposed model facilitaties using typical transaction data to evaluate the effectiveness of relationship marketing activities as well as the impact of past buying behavior on the dynamics of customer relationships and the subsequent buying behavior. My approach to modeling relationship dynamics is structurally different from the models in the existing literature.The model has two following characteristic:Firstly, the exiting models only use trade to scale the relationship between customer and company. In this model I used more factors estimate the relationship, has not the factor numbers limit. Secondly, we can use the relationship states as the standards to subdivision customers dynamically.In addition, I calibrate the proposed model using customer’s buying data provided by an B2C company. This empirical application demonstrates the value of the proposed model in understanding the dynamics of customer-company relationships and predicting buying behavior.更多还原