电信科学 ›› 2019, Vol. 35 ›› Issue (2): 125-133.doi: 10.11959/j.issn.1000-0801.2019040

• 电力信息化专栏 • 上一篇    下一篇

电费回收风险预测的大数据方法应用

赵雅迪1,吴钊2,李庆兵1,陈小峰1,王宝亭1   

  1. 1 北京国网信通埃森哲信息技术有限公司,北京 100032
    2 国网信息通信产业集团有限公司,北京 102211
  • 修回日期:2019-01-20 出版日期:2019-02-01 发布日期:2019-02-23
  • 作者简介:赵雅迪(1993- ),女,北京国网信通埃森哲信息技术有限公司大数据业务咨询顾问,主要研究方向为业务数据分析和数据挖掘。|吴钊(1972- ),男,国网信息通信产业集团有限公司党委组织部(人力资源部)主任、高级工程师,主要研究方向为电力和信息系统。|李庆兵(1975- ),男,北京国网信通埃森哲信息技术有限公司企业管理咨询部副主任、高级信息系统项目管理师,主要研究方向为企业管理咨询与信息化咨询。|陈小峰(1980- ),男,北京国网信通埃森哲信息技术有限公司管理咨询业务总监,主要研究方向为企业管理、精益生产和大数据挖掘。|王宝亭(1981- ),男,北京国网信通埃森哲信息技术有限公司大数据业务咨询顾问,主要研究方向为电力营销业务扩展、计量电费和大数据挖掘。

Application of big data method in forecasting the risk of tariff recovery

Yadi ZHAO1,Zhao WU2,Qingbing LI1,Xiaofeng CHEN1,Baoting WANG1   

  1. 1 State Grid Information &Telecommunication Accenture Information Technology Co.,Ltd.,Beijing 100032,China
    2 State Grid Information &Communication Industry Group Co.,Ltd.,Beijing 102211,China
  • Revised:2019-01-20 Online:2019-02-01 Published:2019-02-23

摘要:

基于电力客户的历史数据,采用客户的基本属性、用电行为、缴费行为、客户信用、行业前景信息等多个维度确定模型所需指标体系。通过指标的相关系数矩阵及信息值(information value,IV)筛选出最终进入模型的指标变量,同时采用最优分组的方法对变量进行分组,并进行证据权重转化(weight of evidence, WOE)。基于处理后的数据,运用逻辑回归算法构建用电客户电费风险预测模型,并依据得到的模型结果量化输出变量标准评分卡表,从而将客户划分为高风险、中风险和低风险用户,为不同的用户采取差异化的营销措施提供依据。

关键词: 电费回收, 逻辑回归算法, 指标体系

Abstract:

Based on the historical data of electricity customers,the model index system was determined according to the customers’ basic attributes,the electricity consumption and the payment behavior,the customers’ credit,the industry prospects’ information and so on.Through the correlation coefficient matrix and the information value of the index,the index variables that enter the model were selected.At the same time,the best grouping method was used to group variables and WOE (weight of evidence) transformation was carried out.Based on the processed data,the logic regression algorithm were used to construct the electricity cost risk forecasting model of the electric customers,and output variable standard score card was quantified according to the model results.Thus the customers were divided into high,middle and low risk users that could provide the basis for taking differential marketing measures to the different customers.

Key words: tariff recovery, logical regression algorithm, index system

中图分类号: 

No Suggested Reading articles found!