基于数据挖掘的运营商流量经营分析与研究

发布时间：2018-03-03 13:13

本文选题：运营商　切入点：流量经营　出处：《南京邮电大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着移动互联网的高速发展,运营商传统的语音、短信业务收入明显下降,各种移动互联网业务快速发展,移动数据流量正急速增长。在流量急速增长的同时,运营商面临着增量不增收的问题,而且有被互联网公司管道化的威胁。在这样的情况下,运营商做好流量经营尤为重要。本文通过分析运营商流量经营的必要性及优缺点所在,通过建立数据挖掘模型,对用户流量数据进行分析研究。以某运营商2015年6月份的南京用户流量数据为基础,采用远程数据采集、本地环境搭建、本地分析用户流量数据的思路,从客户细分、上网偏好划分、关联规则寻找三个方面分析流量经营中的问题:客户细分主要以客户流量行为、消费特征为指标,采取k-means算法对用户进行细分,在结合业务知识的基础上,最终细分为高、中、低三个档次的用户群;上网偏好划分采取高频偏好与普通偏好分离的方法,分别对其聚类,并分别得到4种单偏好群体、1种多偏好群体与9种单偏好群体、8种多偏好群体;然后加入用户终端信息,通过FP-growth算法挖掘它们之间的强规则,得到16条终端、客户档次、上网偏好方面有价值的关联规则,最终得出结论:新闻、视频、财经多偏好用户更倾向发展为低档次客户;高档次客户更喜欢科技、阅读和生活服务类的上网偏好;当用户使用4G制式的手机时,使用华为品牌的置信度为65%,高于苹果品牌的62%。本文的研究为运营商流量经营提供了分析思路,完善系统及操作界面后,可作为运营商在流量经营中进行针对性营销的辅助工具。
[Abstract]:With the rapid development of mobile Internet, the revenue of traditional voice and short message service of operators has decreased obviously, various mobile Internet services have developed rapidly, and mobile data traffic is growing rapidly. At the same time of rapid growth of traffic, Operators are faced with the problem of no incremental revenue increase, and there is a threat of being managed by Internet companies. Under such circumstances, it is particularly important for operators to do traffic management well. This paper analyzes the necessity, advantages and disadvantages of operators' traffic management. Through the establishment of data mining model, the user traffic data are analyzed and studied. Based on the Nanjing user traffic data of a certain operator in June 2015, the remote data collection is adopted and the local environment is built. Local analysis of user traffic data ideas, from customer segmentation, Internet preference division, association rules looking for three aspects of traffic management problems: customer segmentation is mainly based on customer traffic behavior, consumption characteristics as indicators, The k-means algorithm is adopted to subdivide the users, and on the basis of combining business knowledge, it is finally subdivided into three user groups of high, middle and low levels; the division of Internet preference adopts the method of separating high frequency preference from common preference, and clustering them respectively. Four single preference groups, one multi-preference group and nine single-preference groups are obtained, and then the user terminal information is added, and the strong rules between them are mined by FP-growth algorithm, and 16 terminals and customer grades are obtained. The valuable association rules in the aspect of internet preference are concluded: news, video, finance and economics prefer users to develop into low-grade customers, high-grade customers prefer technology, reading and life service. When users use 4G mobile phones, the confidence of using Huawei brand is 65, which is higher than that of Apple brand. Can be used as operators in traffic management targeted marketing tools.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F626;TP311.13

【参考文献】