联系我们

论文查询结果	返回搜索

论文编号：	1049
作者编号：	2120071908
上传时间：	2009/5/29 14:43:19
中文题目：	基于粗糙集的电信业客户流失预测
英文题目：	Telecom churn Forecast Researc
指导老师：	安利平
中文关键字：	组合学习算法 粗糙集 客户流失
英文关键字：	Rough set Ensemble Method Cu
中文摘要：	随着国内电信企业改革的不断进行，我国电信运营业的竞争不断加剧，电信企业对客户的争夺也越来越激烈。这迫使电信经营者必须从其他方面获得竞争优势，他们不仅仅要考虑如何降低运营成本同时还要考虑如何挽留客户，进而保证收入不因竞争加剧而减少。粗糙集技术作为发现数据中知识的方法，近年来受到普遍的关注，并且在银行、医疗、设备故障诊断、入侵防范等诸多方面得到了成功的应用。将粗糙集技术或者其他数据挖掘技术应用于电信行业可以给电信经营者带来实际操作上的指导。其中，客户流失预测分析是通过对以往流失客户的历史数据进行分析，在客户采取离网行为以前发出警示，及时采取相应的措施，减少客户流失的发生。这无疑可以在电信经营者维持稳定收入，降低经营成本方面带来新的竞争优势。 本文研究的目的就是将粗糙集技术应用于电信客户流失预测的领域中，建立电信业客户流失预测模型以辅助电信经营者的经营活动，主要内容包括： 1．客户流失主要问题的研究 总结、概述了客户流失分析中的主要问题，指出这些问题在各方面的内在联系。指出了客户流失分析的定义及其在客户分析中的意义。同时指出客户流失预测与客户流失分析之间的内在关系。 2．基于粗糙集的客户流失预测问题研究框架 总结、综述了数据挖掘在客户分析中的应用，分析了客户流失研究的主要内容，提出了基于粗糙集模型的客户流失问题研究框架，运用粗糙集方法构建电信业客户流失预测模型，并对研究框架中的几大主要部分进行了详细的阐述。 3．基于组合算法和粗糙集模型的客户流失预测模型研究 通过分析客户流失预测问题的数据特点，提出现有方法的不足，并且在属性选择和算法设计两方面进行了讨论，建立基于组合算法和粗糙集理论的电信业客户流失预测模型，取得了很好的效果，并对结果进行了深入的分析与讨论。
英文摘要：	In the process of evolution in our country, competition in the telecommunications industry becomes more and more fierce. Not only telecommunications industry, but also banking and retailing business are facing great pressure from market. They are in the same boat as in customer acquirement, retention and product promotion. In this background they come to seek advantage from other rooms. They have to keep cutting down the running cost while working on making plans to prevent customer churn which guaranty their revenues in competition. Rough set theory is some kind of new technology to discover knowledge from dataset which gains continuous attention these years. It can be used in banking, medical, equipment diagnoses, intrusion prevention and other field. Massive data was produced by telecommunication carriers and it is an deal place to perform data mining. Progress has been achieved in such fields: customer fault detection, customer churn analyses, customer patterns in behavior, and product promotion. On this point of view, rough set technology can be applied in telecommunications, so as other data mining techniques which can support decision making for those carriers. To predict customer churn, history data needs to be analyzed and determines the rules which affect customer behavior. Once applied the predicting model, the customer who will churn in the later period will be checked out before it happens. If necessary action was taken out to stop customer churn, telecommunication service providers will keep his income steady and win advantage in competition. The aim of this research is to apply rough set theory to the field of customer churn prediction in the telecommunications industry. New prediction model is constructed to support decision making in telecommunication carriers’ running activities. New progress was made by combining ensemble method and rough set theory to construct rough set ensemble model based on different cut sets, hoping that improvement in accuracy being achieved. The studying object of this research is a telecommunication company in Tianjin. Based on the statement above, main effort of this research is spend on: (1) Theoretical research on customer churns management Summarize the key points in customer churn management and check out the relationships among these points. Give precise statement of customer churn analyst and show the significance of churn prediction in customer analyst. Study the relationship between churn prediction and customer analyst. (2) Research frame of customer churn prediction based on rough set Summarize the key points of data mining in customer churn analyst and the main issue of churn research. Let out the frame of customer churn prediction based on rough set and construct predicting model. Explain the model in detail. (3) Research of new predicting model based on rough set and ensemble method Check out the specialty of churn prediction in telecommunication industry and summarize existing method. Make discussion both on field selection and model enhancement. Combine ensemble method and rough set theory to construct rough set ensemble model based on different cut sets and check out the result, make discussion and show the direction of further research.
查看全文：	预览  下载(下载需要进行登录)