基于多维向量空间的通信运营指标相似性量化测度研究  

Research on Similarity Measure of Communication Operation Index Based on Multidimensional Vector Space

作　　者：刘世伟[1] LIU Shiwei(China Mobile Group Guangdong Co., Ltd., Dongguan Branch, Dongguan 523000, China)

机构地区：[1]中国移动通信集团广东有限公司东莞分公司,广东东莞523000

出　　处：《移动通信》2019年第3期44-48,53,共6页Mobile Communications

摘　　要：为了提高市场运营分析质量,达到精细化调度市场资源和差异化评价KPI考核体系的目的,通过数据挖掘的方法,针对同一套指标,运用不同的相似性测量方法进行量化测度,分析指标之间相关性以及同一指标在不同方法下的相关性程度。发现当前通信运营指标并不是每个相关系数的运算结果都能给出圆满的解释,难免会出现毫无关联的指标之间存在强相关性,又或者现实中有着强关联的指标却算出毫不相关的结果。由此在制定运营指标体系的时候,分析人员需要加大数据挖掘力度,从更深层次的业务逻辑出发,透过现象看本质,建立一个更加合理的以业务为核心的多层次分析架构。In order to improve the quality of market operation analysis and refine dispatching market resources and differentiated evaluation of KPI assessment system, this paper uses data mining method to quantify the same set of indicators, using different similarity measurement methods, and to analyze the correlation between indicators and the same indicators under different methods. It is found that the perfect explanation does not always hold to the computation results of each correlation coeffi cient for the current communication operation indicators, or there are some strong correlation indicators in the reality but with unrelated results. Therefore, when formulating the operation index system, the analysis personnel should pay more attention to the data mining so that the essence can be seen through the phenomenon and a feasible multi-level analysis framework oriented to business can be established according to the deeper business logic.

关 键 词：相关性分析 数据挖掘 多维向量 余弦距离 皮尔逊相关系数 

分 类 号：TN929.53[电子电信—通信与信息系统]