电信科学 ›› 2021, Vol. 37 ›› Issue (5): 52-63.doi: 10.11959/j.issn.1000-0801.2021113

• 专题:通信与AI融合 • 上一篇    下一篇

基于行为画像的物联网业务保障方法

赵龙刚1, 刘汉生1, 王峰1, 狄爽2   

  1. 1 中国电信股份有限公司研究院,北京 102209
    2 中国电信集团有限公司,北京100032
  • 修回日期:2021-05-10 出版日期:2021-05-20 发布日期:2021-05-01
  • 作者简介:赵龙刚(1977− ),男,中国电信股份有限公司研究院 AI 研发中心高级工程师,主要研究方向为AI能力平台、AI可视化建模、行业应用大数据研究等
    刘汉生(1993− ),男,中国电信股份有限公司研究院 AI研发中心工程师,主要研究方向为网络智能化运维、威胁情报等
    王峰(1979− ),男,博士,中国电信股份公司研究院教授级高级工程师,长期从事云计算、大数据、人工智能等新兴信息技术领域的技术研发和产品创新工作
    狄爽(1992− ),女,现就职于中国电信集团公司云网运营部(大数据和AI中心)智能云网调度运营中心(ICNOC),主要研究方向为物联网业务运营分析优化、网络运维等
  • 基金资助:
    国家重点研发计划基金资助项目(2019YFB1802501);国家重点研发计划基金资助项目(2019YFB1802504)

IoT business guarantee method based on behavioral portrait

Longgang ZHAO1, Hansheng LIU1, Feng WANG1, Shuang DI2   

  1. 1 Research Institute of China Telecom Co., Ltd., Beijing 102209, China
    2 China Telecom Group Co., Ltd., Beijing 100032, China
  • Revised:2021-05-10 Online:2021-05-20 Published:2021-05-01
  • Supported by:
    The National Key Research and Development Program of China(2019YFB1802501);The National Key Research and Development Program of China(2019YFB1802504)

摘要:

物联网终端具有客户基数大、生产厂商多、应用场景复杂的特点,在日常维护过程中存在质差标准难以统一、定位定段困难的问题。针对上述现象提出一种基于行为画像的业务保障方法。首先基于关键指标分布特征构造企业质差指纹模型,借鉴统计学习中均值漂移聚类的思想,实现质差指标体系的准确搭建。然后针对调测终端与质差终端难区分、弱覆盖终端难识别等问题,构建了单客户质差行为画像,有效保证了模型的准确性。最后在现网环境进行了试点和分析,为物联网业务保障提供借鉴和参考。

关键词: 物联网, 质差识别, 根因分析

Abstract:

IoT terminals have the characteristics of large user base, many manufacturers and numerous scenarios.It is difficult to unify the standard of poor quality and to locate the segment in the routine maintenance process.Aiming at the above phenomenon, a business guarantee method based on behavior portrait was proposed.Firstly, based on the distribution characteristics of key indicators, a fingerprint model of enterprise quality deficit was constructed, and the idea of mean shift clustering in statistical learning was used to realize the accurate construction of quality deficit index system.Then, to solve the problem that it was difficult to distinguish between the measurement terminal and the poor quality terminal, and it was difficult to identify the weak coverage terminal, a single user poor quality behavior portrait was constructed to effectively ensure the accuracy of the model.Finally, the pilot and analysis were carried out in the current network environment to provide reference for the IoT business guarantee.

Key words: internet of things, poor quality identification, root cause analysis

中图分类号: 

No Suggested Reading articles found!