基于大数据的4G资费研究与分析

发布时间：2018-01-30 18:04

本文关键词： 4G资费大数据数据挖掘资费套餐　出处：《天津商业大学》2017年硕士论文　论文类型：学位论文

【摘要】：截止2016年7月,我国总人口达到13.83亿人,移动电话用户总数达到13.21亿户,其中4G用户总数达到7.14亿户。2G、3G和4G移动用户总数量已接近饱和,单纯增加4G移动用户数量的增量式增长模式已经不能满足移动运营商的需求,必须将挖掘已在网2G、3G用户到4G用户经营增长模式作为移动运营商的发展重点,这对运营商的业务增长和我国通信领域的经济建设有着重要意义。本文针对某省公司电信运营商的用户数据,运用决策树、随机森林、AdaBoost、支持向量机和K-means等算法,从“挖掘潜在4G用户模型”和“4G用户资费特征模型”两方面对4G用户的特征及资费进行分析。首先,通过探索性分析对4G用户的基本特征(性别、年龄等)及消费特征(ARPU、DOU、MOU等)进行宏观了解。然后,利用数据挖掘技术对用户数据进行分析,在“挖掘潜在4G用户模型”中,支持向量机算法的准确率有着巨大优势,通过对变量进行对数变换,进一步提高了准确率;在“4G用户资费特征模型”中,通过对用户DOU、MOU两个变量进行聚类分析,对用户整体消费特征进行把握,得出“均衡型”、“数据流量驱动型”、“语音通话驱动型”三大类用户,并通过对不同性别和总体聚类特征对比,得出k=13时,对用户资费的分类程度最好且成本较低,根据聚类分析结果为运营商制定资费套餐时提出意见或建议。
[Abstract]:By July 2016, the total population of China reached one billion three hundred and eighty-three million people, the total number of mobile phone users reached one billion three hundred and twenty-one million, of which the total number of 4G subscribers reached 714 million. The total number of 3G and 4G mobile users is nearly saturated, and the incremental growth mode of increasing the number of 4G mobile users is no longer able to meet the needs of mobile operators, must be excavated in the network 2G. 3G users to 4G users business growth model as a mobile operator development focus. This is of great significance to the business growth of operators and the economic construction in the field of communications in China. This paper applies decision tree and random forest AdaBoost to the customer data of telecom operators in a province. Support vector machine and K-means algorithm, from "mining potential 4G user model" and "4G user tariff characteristics model" two aspects of the 4G user characteristics and tariff analysis. Through exploratory analysis of 4G users' basic characteristics (gender, age, etc.) and consumption characteristics of ARPUUDOUMOU, etc. Using data mining technology to analyze user data, in "mining potential 4G user model", the accuracy of support vector machine algorithm has a huge advantage, through the logarithmic transformation of variables. The accuracy rate is further improved; In the "4G user tariff characteristic Model", through clustering analysis of two variables of user DOUUMOU, the author grasps the overall consumption characteristics of users, and obtains "balanced" and "data flow driven". "Voice call driven" three categories of users, and through the comparison of different gender and overall clustering characteristics, it is concluded that k = 13:00, the best classification of user fees and lower cost. According to the cluster analysis results for operators to develop tariff packages put forward comments or suggestions.
【学位授予单位】：天津商业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F626

【参考文献】