摘 要:随着全市家庭宽带市场的饱和,宽带用户新增速度放缓,保有存量用户、控制用户离网业已成为促进宽带市场发展的重要举措。文章对家庭宽带离网用户特征进行研究,基于 lightGBM、XGBoost、RandomForest 三类集成学习的决策树算法,使用 PyCharm 软件构建家庭宽带离网用户预警模型,输出预离网用户供业务人员进行挽留,模型应用后,宽带月离网用户百分比从 0.76% 下降至 0.35%,预计全年可挽回预离网用户 7 776 户,保有客户价值 101.1 万元。
关键词:离网;大数据;预测;量化;宽带用户
DOI:10.19850/j.cnki.2096-4706.2021.15.022
中图分类号:TP311 文献标识码:A 文章编号:2096-4706(2021)15-0085-04
Research on the Construction of Home Broadband Off-network User Early Warning Model
ZHANG Jing, HOU Xiaojing
(Shuozhou Branch of China Mobile Communications Group Shanxi Co., Ltd., Shuozhou 036002, China)
Abstract: With the saturation of the home broadband market in the whole Shuozhou city, the growth rate of broadband users has slowed down. Retaining existing users and controlling user off-network have become important measures to promote the development of the broadband market. This paper studies the characteristics of home broadband off-network users, based on the decision tree algorithm of integrated learning of lightGBM, XGBoost and RandomForest, PyCharm software is used to construct home broadband off-network users early warning model, which outputs pre off-network users for business personnel to retain. After the application of the model, the percentage of monthly broadband off-network user drops from 0.76% to 0.35%. It is expected that 7 776 pre off-network users can be retained throughout the year, keeping a customer value of 1.101 million yuan.
Keywords: off-network; big data; prediction; quantification; broadband user
参考文献:
[1] 王远征,吴峰,夏明镜,等 . 电信宽带用户离网大数据预测实例 [J]. 电信技术,2016(10):83-87.
[2] 云晴.大数据实践重要影响因素 [J].中国电信业,2017(8):57-59.
[2] 卢光跃,张宏建,闫真光,等 . 基于特征选择和 SVM 的电信客户离网预测 [J]. 西安邮电大学学报,2019,24(2):21- 25.
[3] 陈小凤 . 宽带用户维系新思路 [J]. 电子技术与软件工程,2018(14):14.
[4] 谷娜 . 基于 AI 和大数据的宽带服务支撑体系研究 [J]. 通信与信息技术,2021(4):81-82.
[5] 云晴.大数据实践重要影响因素 [J].中国电信业,2017(8):57-59.
作者简介:张靖(1986—),男,汉族,山西朔州人,中级工程师,硕士研究生,研究方向:神经网络;侯晓晶(1987—),女,汉族, 山西省运城人,中级工程师,硕士研究生,研究方向:数据挖掘。