SIA OpenIR  > 工业信息学研究室  > 先进制造技术研究室
基于群体智能的数据挖掘方法及在CRM中应用研究
Alternative TitleResearch on Data Mining Methods and Applications in CRM Based on Swarm Intelligence
金鹏1,2
Department先进制造技术研究室
Thesis Advisor朱云龙
ClassificationTP18
Keyword数据挖掘 群体智能 蚁群算法 转移模式分析 客户关系管理
Call NumberTP18/J67/2008
Pages102页
Degree Discipline机械电子工程
Degree Name博士
2008-01-05
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract随着数据采集技术的不断发展、数据库系统的日益完善,人们已经从日常的生产制造、科学研究、商务运作、消费行为等各行各业的活动中积累了海量的原始数据,如何从纷繁复杂的数据中发现有价值的知识,是目前关于数据分析处理的理论研究和实际应用中亟待解决的问题。数据挖掘技术作为一种有效的知识发现手段,得到了广泛的研究和应用。由于实际应用中数据源的丰富性和复杂性,需要根据数据源的特点和应用领域的特征及要求来选择合适的数据挖掘功能和算法,因此新算法的研究具有理论和应用价值。 群体智能是受到群居昆虫群体和其它动物群体集体行为的启发而产生的算法和解决方案的总称,是一类新兴的启发式搜索算法。由于其在解空间中具有鲁棒性和柔性良好的解搜索能力,因此在旅行商问题、二次分配、车间调度、序列求序、图形着色、网络路由以及群体机器人协同等问题中得到了广泛应用。 本文对基于群体智能的数据挖掘方法进行了研究,改进和提出了基于群体智能的分类和预测方法以及聚类分析方法。根据群体智能中个体行为与客户关系管理领域中客户行为的相似性,提出了基于群体智能的客户转移模式分析方法,并针对汽车行业营销管理应用构建了客户转移模式分析仿真平台。在上述研究的基础上提出了客户关系管理中基于群体智能的数据挖掘应用体系,对客户关系管理所涉及的客户细分、客户流失预测、客户价值分析、交叉销售和纵深销售以及客户转移模式分析等业务应用提供决策支持。本文主要研究内容及成果如下。 在基于群体智能的数据挖掘方法研究方面: 对基于蚁群优化算法的分类规则挖掘方法进行了详细的研究和分析,发现现有算法在信息素更新策略、对初始项的依赖、离散化方法选择以及参数的设定等几方面存在不足,针对上述问题,提出了基于蚁群优化算法的分类规则挖掘方法ACO-Classifier,算法中采用了改进的信息素更新策略、多种群并行策略、考虑误分类代价的离散化方法,并对算法中的参数进行步进式调整,从而在算法预测准确度和规则简洁性方面取得了更好的性能; 根据蚁堆聚类形成的原理,提出了用于数据挖掘聚类分析的Ant-Cluster算法,在该算法中引入了具有不同运动速度的多蚂蚁种群,给出了以随机速度运动蚂蚁的速度计算方法,解决了孤立点数据对象造成的蚂蚁锁定问题,与常用的划分聚类方法k-means算法相比,不必预先制定簇的个数,并对数据进行了降维,将多维数据放到二维空间中,可以构造任意形状的簇,同时由于只需要计算与蚂蚁临近数据对象间的相似性,而不需要遍历所有数据对象,因此大大减少了计算量; 针对当前客户关系管理中对动态的客户生命周期价值和客户行为模式分析的需求,提出了客户转移模式的概念,研究了客户转移模式分析的流程,根据群体智能中个体行为与客户关系管理领域中客户行为的相似性,提出了基于群体智能的客户转移模式分析方法,将单一的客户数据作为独立个体,给出了其在规则集中搜索的信息素更新策略及项搜索策略,从而挖掘出客户转移模式,使企业能够及时准确地掌握客户需求及其消费模式的变化趋势并制定出相应的市场营销策略。 在基于群体智能的数据挖掘方法应用方面: 在客户转移模式分析方法研究的基础上,对客户数据动态分析方法进行进一步研究,从客户行为模式、客户生命周期价值及客户转移成本的角度,对客户在不同营销策略下的转移趋势进行预测分析和仿真评估,从而将营销策略的事后控制变为事前预测,可有效地实现对营销策略的及时控制和修正,并针对汽车行业营销管理应用构建了客户转移模式分析仿真平台,对汽车行业的营销策略的制定与评估具有重要的指导意义; 对制造业客户关系管理中的数据挖掘应用进行了系统的分析,提出了客户关系管理的业务应用、评价指标以及数据挖掘功能间的映射关系,分析了数据挖掘在客户细分、客户流失预测、客户价值分析、交叉销售和纵深销售、客户转移模式分析等业务领域的应用,根据这些应用间的相互联系,采用基于群体智能的数据挖掘方法,构建了系统的制造业客户关系管理中的数据挖掘应用体系,同时该应用体系可扩展到其它领域和行业的客户关系管理中。 总之,本文在对基于群体智能的数据挖掘方法进行研究的基础上,将算法研究成果应用于客户关系管理,构建了客户转移模式分析仿真平台及针对客户关系管理各种业务应用的数据挖掘应用体系。本研究对于数据挖掘方法及分析型客户关系管理的研究和应用具有促进和指导作用。
Other AbstractWith the development of data acquisition technology and database system, huge data sets from many domains, such as manufacture, scientific research, business operation, consumption behaviour, and so on, are cumulated. It is emergent and important to discovery valuable knowledge from complicated and huge data sets. Data mining, as an effective approach for knowledge discovery, has been researched and applied widely. Due to diversity and complexity of data source in practical application, the appropriate functions and algorithms of data mining should be chosen according to the characteristics of data source and application domain. New algorithms for data mining are valuable for both theory and application research. Swarm intelligence is a general designation of algorithms or distributed problem-solving devices inspired by the collective behavior of social insect colonies and other animal societies. It has some advantages and characteristics, such as self-adaptation, self-government, and parallel computing, etc. Due to the good robusticity and flexibility for searching solution, it has been applied in the traveling salesman problem (TSP), quadratic assignment problem, graph coloring, job-shop scheduling, sequential ordering, vehicle routing, network routing, and multi-robot cooperation, etc. Data mining methods based on swarm intelligence are researched in this thesis. The classification and predication method and clustering analysis method based on swarm intelligence are improved and presented. According to the comparability of individual behaviour in swarm intelligence and customer behaviour in customer relationship management (CRM), customer change model analysis method based on swarm intelligence is presented. The customer change model analysis simulation platform is built focusing on automobile industry. Based on these researches, the customer relationship management prototype system based on swarm intelligence is established. This system provides decision supports for business applications of customer relationship management, such as customer segmentation, customer churn prediction, customer value analysis, cross-selling/up-selling, and customer change model analysis. The researches of this thesis include the contents and results as follows. The classification rules mining methods based on ant colony optimization (ACO) are researched and analyzed. The shortages of existing algorithms are discovered, including pheromone updating strategy, dependence on initialization items, choice of discretization method, and setting up of parameters. To solve these problems, ACO-Classifier algorithm is proposed. The improved strategy of pheromone updating in MAX-MIN Ant System is adopted, multi-population parallel strategy is proposed, the cost-based discretization method is used, and parameters in the algorithm are adjusted step by step. With these improvements, performance of the algorithm is advanced. The results of experiments illuminate that the ACO-Classifier algorithm has better performance on predictive accuracy and simplicity of rules. According to the principle of ant clustering, a clustering algorithm for data mining based on swarm intelligence called Ant-Cluster is proposed. Ant-Cluster algorithm introduces the concept of multi-population of ants with different speed, gives the calculating method of ants with random speed, and adopts fixed moving times method to deal with outliers and locked ant problem. Comparing with the k-means algorithm, a wildly used partition clustering method, Ant-Cluster algorithm can get clustering results effectively without giving the number of clusters. The dimensions of data are reduced through putting multi-dimension data into 2-dimension grid. Another advantage to do so is that clusters with any shape can built. The calculating amounts are decreased because only neighbour data of an ant need be calculated. Understanding and adapting to changes of customer behavior is an important aspect of surviving in a continuously changing market environment for a modern company. With the development of new business models and continuous change of customer demand and behavior model, the dynamic analysis of customer data and customer relationship management have to face new challenges. The concept of customer change model mining is introduced and its process is analyzed. A customer change model mining method based on swarm intelligence is presented to discover the change of rules not from the aspect of rule structure but the change of customers, and the strategies of pheromone updating and items searching are given. This customer change model mining method can help company to understand customer demands and consumption model change and make appropriate market strategy in time. Based on the research of customer change model analysis method, the dynamic analysis method of customer data is researched farther. According to customer behaviour model, customer lifetime value, and customer transfer cost, customers change trend under different marketing strategy are predicated and evaluated. The post-control of marketing strategy becomes pre-predication so that marketing strategy can be controlled and modified in time. The customer change model analysis simulation platform is built focusing on automobile industry. This system is significant for marketing strategy making and evaluating in automobile industry. Applications of data mining can support manufacture CRM effectively, and a systematical architecture of data mining has been needed for every aspects of manufacture CRM. The mapping relationships among business applications, evaluating indicators and data mining methods are presented. Based on the mapping relationships, these business applications, including customer segmentation, customer churn prediction, customer value analysis, cross-selling/up-selling, and customer change model analysis, are studied. Data mining methods based on swarm intelligence including clustering analysis, classification rules mining, and customer change mining are adopted to build applications architecture of data mining in manufacture CRM. The architecture can be extended to CRM of other domain. In conclusion, this thesis researchs data mining methods based on swarm intelligence, and the results of algorithm research are applied in customer relationship management. The customer change model analysis simulation platform and customer relationship management prototype system based on swarm intelligence are established. Thus it is undoubtedly to say that this thesis is of certain significance and realistic application value for data mining methods and CRM research.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/132
Collection工业信息学研究室_先进制造技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院研究生院
Recommended Citation
GB/T 7714
金鹏. 基于群体智能的数据挖掘方法及在CRM中应用研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2008.
Files in This Item:
File Name/Size DocType Version Access License
10001_20031801470304(2378KB) 开放获取--Application Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[金鹏]'s Articles
Baidu academic
Similar articles in Baidu academic
[金鹏]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[金鹏]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.