摘要
采用自监督方式检测异常小区、无监督方式聚类问题类型、可视化标注标签的数据挖掘思路,得到带标签的异常小区问题类型数据集。不仅标签数据集可直接应用到网络运营运维领域,而且挖掘问题类型数据集的思路和方法论可广泛推广到其他领域。另外,借助卷积神经网络深度学习算法,将大量的专家经验模型化,构建智能优化引擎,挖掘出人工不易捕捉到的深度信息,更加准确高效地发现网络问题,可扩展性强。
Using self-supervised method to detect abnormal cells,unsupervised method to cluster problem types and visually labeled data mining ideas,the abnormal cell problem type data sets with labels are obtained. Not only label data set can be directly applied to the field of network operation and maintenance,but also the idea and methodology of mining problem type data set can be widely extended to other fields. In addition,with the help of convolutional neural network deep learning algorithm,a large number of expert experience is modeled,and an intelligent optimization engine is constructed to mine the depth information which is not easily captured by human,so that network problems can be found more accurately and efficiently with strong scalability.
作者
王勇
滕祖伟
周杰华
肖波
赵根
Wang Yong;Teng Zuwei;Zhou Jiehua;Xiao Bo;Zhao Gen(China United Network Communications Group Co.,Ltd.,Beijing 100033,China;China Unicom Hubei Branch,Wuhan 430020,China)
出处
《邮电设计技术》
2019年第11期11-15,共5页
Designing Techniques of Posts and Telecommunications
关键词
深度学习
异常小区
问题检测
问题分类
Deep learning
Abnormal cell
Problem detection
Problem classification