基于数据挖掘的电信客户流失预测研究

发布时间：2018-04-22 23:35

本文选题：数据挖掘 + 客户流失预测　；参考：《西安电子科技大学》2012年硕士论文

【摘要】：随着计算机技术特别是数据库技术的广泛应用，各行各业都积累了海量的数据。为了消除“数据爆炸但知识贫乏”的现象，数据密集型企业越来越认可数据挖掘的重要性。基于数据挖掘的电信客户流失预测这项研究，是在数据仓库和数据挖掘技术迅速发展的基础上，针对电信企业客户关系管理的迫切需要，为消除客户流失给运营商带来的不利影响而提出的。本文的工作基于Q省移动“客户流失预测系统”项目背景，将数据挖掘技术应用到电信企业的客户流失预测中。以Q省移动客户数据、业务数据为依据，按照商业理解、数据理解、数据准备、建立模型、模型评价、模型发布的步骤，利用统计分析软件SPSS及数据挖掘工具Clementine设计和建立了电信客户流失预测的综合模型。本文首先介绍了数据挖掘理论及分类预测算法，并详细描述了生存分析理论及比例风险模型。在建模过程中，重视数据质量，进行了有效的数据清洗、转换、探索工作，，处理了不平衡数据集，并在业务经验及属性约简的基础上建立了流失预测指标体系。最终建立了决策树、神经网络、logistic回归以及生存分析Cox模型，并对模型进行了多项指标的评估。在维系挽留工作中，初步分析了客户流失原因并评定了客户价值，提出针对不同客户进行因时因地的有针对性的维系挽留策略，减少挽留成本并提高挽留的成功率。本文把数据挖掘理论与实际项目相结合，建立了基于数据挖掘技术的电信客户流失预测综合模型。理论研究上对分类模型及Cox模型的构建具有指导意义；应用的结果表明所建立的模型能够给决策人员提供有价值的预测信息并给出相应的解决方案。
[Abstract]:With the wide application of computer technology, especially database technology, huge amounts of data have been accumulated in various industries. In order to eliminate the phenomenon of "data explosion but poor knowledge", data-intensive enterprises increasingly recognize the importance of data mining. The research of telecom customer churn prediction based on data mining is based on the rapid development of data warehouse and data mining technology, aiming at the urgent need of customer relationship management in telecom enterprises. In order to eliminate the negative impact of customer drain on operators. Based on the project background of Q province mobile customer churn prediction system, this paper applies data mining technology to customer churn prediction of telecom enterprises. Based on Q province mobile customer data, business data, according to business understanding, data understanding, data preparation, modeling, model evaluation, model release steps, Using the statistical analysis software SPSS and the data mining tool Clementine, a comprehensive model of telecom customer churn prediction is designed and established. This paper first introduces the theory of data mining and classification and prediction algorithm, and describes the survival analysis theory and proportional risk model in detail. In the process of modeling, we attach importance to data quality, carry out effective data cleaning, transformation, exploration, deal with unbalanced data sets, and establish a loss prediction index system on the basis of business experience and attribute reduction. Finally, the decision tree, neural network logistic regression and survival analysis Cox model are established, and several indexes are evaluated. In order to reduce the cost of retention and improve the success rate of retention, this paper analyzes the reasons of customer turnover and evaluates the value of customers, and puts forward a targeted retention strategy aimed at different customers in order to reduce the cost of retention and improve the success rate of retention. This paper combines the theory of data mining with practical projects, and establishes a comprehensive model of telecom customer churn prediction based on data mining technology. The theoretical research is of guiding significance to the construction of classification model and Cox model, and the application results show that the established model can provide valuable prediction information and corresponding solutions for decision makers.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP311.13;F274;F626

【相似文献】