数据挖掘在电信手机用户识别中的应用

发布时间：2018-03-04 00:19

本文选题：数据挖掘　切入点：高端用户　出处：《浙江工商大学》2013年硕士论文　论文类型：学位论文

【摘要】：随着通信行业的发展、手机的普及,我国电信行业的竞争越来越激烈。电信行业的经营理念由原来的以产品为中心转变为以客户为中心。电信行业通过扩大手机用户来提高自己的业务已经非常困难。目前,对运营商来说最重要的是保留用户,尤其是高价值用户,避免用户流失。保留用户的首要前提是了解用户。首先,本项目研究如何识别电信公司高端用户。根据电信公司业务人员对高端用户的商业理解大概确定高端用户应该满足的条件,在此基础上使用数据挖掘方法具体确定哪些用户为高端用户,在此基础上分析高端用户应该满足的特征。在分析高端用户时,本项目从两个角度进行了研究：业务角度和统计角度。最后合并两种方法得到的结果确定最终的高端用户。其次,本项目研究流失用户识别。随着电信行业竞争加剧,没有一个运营商可以肯定其用户不会转向其他运营商。因此,预测哪些用户将会流失是运营商保留用户的前提。本项目在理解业务的基础上,先对流失用户进行界定,然后用决策树算法进行建模,并对算法CR、 QUEST、C5.0和基于Boosting的C5.0得到的结果进行比较,从中选择基于Booting的C5.0作为本项目最终使用的模型。最后,从被预测为将会流失的用户中选出高端用户,作为运营商挽留的重要用户。在此基础上,分析每类被预测为流失的高端用户挽留措施。
[Abstract]:With the development of the communication industry, the popularity of mobile phones, The competition of telecommunication industry in our country is more and more intense. The management idea of telecommunication industry has changed from product center to customer center. It is very difficult for telecom industry to improve its business by expanding the number of mobile phone users. At present, The most important thing for operators is to keep users, especially high-value users, to avoid the loss of users. First of all, this project studies how to identify high-end users of telecom companies. According to the business understanding of high-end users of telecom companies, we can determine the conditions that high-end users should meet. On this basis, we use data mining method to determine which users are high-end users, and then analyze the characteristics that high-end users should satisfy. This project has carried on the research from two angles: the business angle and the statistical angle. Finally, the result of combining the two methods to determine the final high-end users. Secondly, this project studies the loss of user identification. As competition in the telecommunications industry intensifies, no operator can be sure that its users will not turn to other operators. Predicting which users will be lost is a prerequisite for operators to retain users. Based on the understanding of the business, the project defines the lost users first, and then uses the decision tree algorithm to model the users. The results obtained from the algorithm CR-QUESTC5.0 and C5.0 based on Boosting are compared, and C5.0 based on Booting is selected as the final model of this project. Finally, the high-end users are selected from the users who are predicted to be lost, as the important users to be retained by the operators. On this basis, the retention measures of each type of users are analyzed.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP311.13;F626

【参考文献】