【作者】 罗彬；

【导师】 邵培基；

【作者基本信息】 电子科技大学 ， 管理科学与工程， 2010， 博士

【摘要】 电信客户流失问题不是一个单纯的客户挽留问题,而是一个涉及运营商、客户、政府、技术等多因素交叉影响的复杂系统;在电信客户流失预测中还存在着数据来源众多、数据属性关系复杂、类别数量非对称分布等特点;在电信客户流失挽留中不仅存在影响客户流失与保持的各种效应,而且电信企业必须综合考虑企业内部的挽留资源限制和企业外部的竞争对手反应等条件。而现有关于客户流失分析研究方面还缺乏一套科学的、系统的理论框架和方法体系,现有基于单模型客户流失预测方法也不能完全满足应用需要,现有基于策略概述和定性分析模式的电信客户流失挽留研究对电信企业制定科学的挽留策略指导作用不大。在这种背景下,探索和研究一套新型的电信客户流失分析的理论框架和方法体系,构建一类高效的电信客户流失预测模型和科学的电信客户流失挽留模型将具有重要的理论意义和实践价值。本文主要基于电信客户流失问题本质特征,研究电信客户流失分析理论框架和方法体系,在此基础上围绕提升电信客户流失预测能力和优化电信客户流失挽留策略等目标,展开了一系列电信客户流失预测与挽留研究。首先,电信客户流失管理问题是一个复杂问题,而目前仍缺乏成熟理论指导管理实践。本文在现有解决复杂系统问题的相关思想(综合集成思想、模型集成思想、系统动力思想)启发下,提出了一套基于多模型优化集成(Mutiple Models Optimized Integration,MMOI)的电信客户流失分析理论框架和方法体系。该框架由电信客户流失预测分析模块和电信客户流失挽留分析模块构成,前者主要通过对多个预测子模型的优化并联来实现预测性能的提高,后者主要通过把复杂问题进行化整为零的方式分别建模处理,最后将多个子模型串联集成以求对电信客户流失挽留这个复杂问题的有效解决。研究结论表明,所提出的多模型优化集成(MMOI)的电信客户流失分析理论框架和方法体系可以从理论高度科学指导电信客户流失问题的有效解决。其次,电信客户流失预测问题具有数据来源众多,数据属性关系复杂、类别数量非对称等特点,现有研究大多基于单模型的预测模式已经不能满足电信企业对客户流失精确预测要求。因此,在第二章提出的多模型优化集成(MMOI)的电信客户流失分析理论框架和方法体系指引下,结合不同集成目标(基于预测精度和基于预测收益)、不同集成方式(线性集成、非线性集成、动态集成)、不同集群智能优化技术(人工蜂群算法、人工鱼群算法、人工蚁群算法、粒子群算法、遗传算法)等构建了一系列基于多模型优化集成(MMOI)的电信客户流失预测模型。研究结果表明,所提出的基于多模型优化集成(MMOI)的电信客户流失预测模型不仅预测性能高于普通单模型,而且预测结果也较为稳定。再次,电信客户流失挽留问题具有影响效应内在关系复杂、客户保持意愿呈现动态变化、电信企业内外部限制条件复杂等特点,现有研究大多基于策略概述和定性分析模式已经不能科学指导电信企业展开有效的客户挽留活动。因此,在第二章提出的多模型优化集成(MMOI)的电信客户流失分析理论框架和方法体系指引下,首先针对电信企业挽留资源有限的实际情况,构建了基于预算限制下的电信客户流失挽留分析模型,该模型主要从系统动力学角度详细分析了影响客户保持的三种效应(挽留激励效应、自然流失效应、口碑传播效应)及效应影响系数,再依据客户保持率推导出了客户挽留周期和客户挽留价值计算公式,据此建立了单个客户挽留模型和一对一客户挽留模型;最后针对电信企业挽留策略可能会引发竞争对手反击的实际情况,构建了基于竞争对手反击效应的电信客户流失挽留分析模型,该模型也是从系统动力学视角详细分析了影响客户保持的四种效应(挽留激励效应、自然流失效应、口碑传播效应、对手反击效应)及效应影响系数,再依据客户保持率推导出了客户挽留周期计算公式,以及三种情况下(有竞争反击效应、无竞争反击效应、竞争反击效应不确定)的客户挽留价值模型和客户挽留策略模型。研究结论表明,所提出的基于多模型优化集成(MMOI)的电信客户流失挽留模型能够科学指导电信客户流失挽留决策。最后,对前面所提出的多模型优化集成(MMOI)的电信客户流失分析理论框架和方法体系、基于多模型优化集成(MMOI)的电信客户流失预测模型和电信客户流失挽留模型从理论价值、管理实践及其特点等方面做了分析和评述。更多还原

【Abstract】 Customer churn is not a simple problem whose object is to retain customers, but a complex system involving many factors, such as operators, customers, government and technology, which influence each other. In customer churn prediction, there exist many characters, such as data sources being numerous, data attributes relationship being complex, the number of categories being asymmetric, and so on. In customer churn retention in telecom, there exist various kinds of effects influencing customers churn and customers retention, meanwhile, the telecommunication enterprises must consider the conditions including the limitation of the internal resourses and the response of the external rivals. However, the existing researches on customer churn still lack of a set of scientific, system theory and method; the existing methods for customer churn prediction based on single models cannot completely meet application needs; the existing researches on customer churn retention in telecom based on a summary of strategy and qualitative analysis cannot give much guidance for telecommunication enterprises when making scientific retention stragies. Therefore, it is of important theoretical and practical contribution to explore and study a new theory frame and method system for analyzing customer churn in telecom and to construct a kind of efficient predition models and a kind of scientific retention models for customer churn in telecom. Based on the substantive characteristics of customer churn in telecom, the theory framework and method system for analyzing customer churn in telecom are studied. Then, around the targets of lifting the ability of predicting customer churn and optimizing strategies of customer retention in telecom, a series of researches are launched for customer churn prediction and retention in telecom.First of all, customer churn management in telecom is a complicated problem, and there is little mature theory to guide management practice. Inspired by the existing ideas to solve the problems of complex systems (integrated thoughts, the thoughts of model integration, and the thought of system dynamics), the theory framework and method system are presented to analyze customer churn in telecom based on Multiple Models Optimized Integration (MMOI). The framework consists of customer churn prediction analysis module and customer churn retation analysis module in telecom, the former mainly increase the prediction performance of the system through paralleling the multiple prediction sub-models by optimization, and the latter solves effectively the complex problem of customer churn retention in telecom mainly by separating the complex problem into pieces and modeling, respectively, then integrating the models in series. Research results show that the theory framework and method system analyzing customer churn in telecom based on MMOI proposed in this paper can guide the customer churn problem effectively solved from theoretical height.Secondly, the problem of customer churn prediction has the features of data sources being numerous, the relationship among data attributes being complicated, and the number of the categories being asymmetrical, however, the existing researches based on single prediction model cannot satisfy accurate requirements when predicting the customer churn in telecom. Therefore, guided by the theory framework and method system analyzing customer churn in telecom based on MMOI proposed in the second chapter, combining different integration goals (based on the prediction precision and forecast earnings), different integration ways (linear integration, nonlinear integration, and dynamic integration), different intelligent optimization techniques (Artificial Bee Colony Algorithm, Artificial Fish Swarm Algorithm, Artificial Ant Colony Algorithm, Particle Swarm Optimization Algorithm, and Genetic Algorithm, etc.), we construct a series of customer churn prediction models in telecom based on MMOI. Researches show that not only predict performance of the customer churn prediction model in telecom based on MMOI excels common single models, but also the prediction results is more stable.Thirdly, the problem of customer churn retention has the features of internal relations among the effects being complicated, intend of customer retention changing dynamically, and the limitation on internal and external conditions of the telecommunication enterprises being complicated, however, the existing researches are mostly based on strategies summary and qualitative analysis and cannot scientifically guide telecommunication enterprise to retain the customers effectively. Therefore, guided by the theory framework and method system analyzing customer churn in telecom based on MMOI proposed in the second chapter, based on the actual situation of the limited resourses used to retain customers in telecommunication enterprises, we construct customer churn retention analysis model in telecom based on limited budgets. This model analyzes three effects (retention incentive effect, nature churn effect and word-of-mouth spread effect) affecting customer retention and the coefficients of the effects in detail from the angle of system dynamics, derives the formula calculating customer retention cycle and customer retention values according to customer retention rate, and then establishes the model of single customer retention and the model of one-on-one customer retention. On account of the actual situation of that retention strategy of telecommunication enterprises may cause competitors to counterattack, we construct a customer churn retention analysis model in telecom on basis of rival counterattack effect. The model analyzes four effects (retention incentive effect, nature churn effect, word-of-mouth spread effect, and rival counterattack effect) affecting customer retention and the coefficients of the effects in detail from the angle of system dynamics, derives the formula calculating customer retention cycle according to customer retention rate, and then establishes the customer retention values models and the customer retention strategies models under three condtions (having competition effect, having no competition counterattack effect, and competitive counterattack effect being uncertain). Research results show that the customer churn retention model can guide scientifically the telecommunication enterprises to establish customer retention strategies.Finally, from the angle of theory contribution and management practice, we analyze and evaluate the theory framework and method system of customer churn analysis in telecom based on MMOI, the customer churn prediction models and the customer churn retention models in telecom based on MMOI, their characters, and so on.更多还原