基于时间序列建模的电力营销客户交易行为分析  

Transaction behavior analysis for power marketing customer based on time series modeling

作　　者：方正云 杨捷 杨泓 廖洁萍 段明明 FANG Zheng-yun;YANG Jie;YANG Hong;LIAO Jie-ping;DUAN Ming-ming(Yunnan Power Grid Co.Ltd.,China Southern Power Grid,Kunming 650011,China)

机构地区：[1]中国南方电网云南电网有限责任公司,昆明650011

出　　处：《沈阳工业大学学报》2020年第2期127-131,共5页Journal of Shenyang University of Technology

基　　金：国家科技重大专项项目(2017YFB213827);南方电网科技立项项目(YNKJQQ00000275).

摘　　要：针对电力系统对客户服务的效率低、针对性差的问题,提出了一种采用云模型、离散余弦变换和反向传递神经网络的客户分类算法.通过云模型提取价格敏感度、投诉率、欠费率、销售变现天数以及忠诚度的变化波动情况的云测度,结合余弦变换计算得到特征向量,输入反向传递神经网络训练,从而得到自动客户分级识别的模型.结果表明,提出的模型可以有效实现客户分级以及销售状况评估,其平均检测精确度可达95%以上.In view of low efficiency and poor pertinence of power system for customer service,a customer classification algorithm based on cloud model,discrete cosine transformation and reverse transfer neural network was proposed. The cloud measure of fluctuating situation in price sensitivity,complaint rate,arrears rate,sales cash-in days and loyalty were extracted by a cloud model;eigenvectors were calculated and obtained by cosine transformation;the reverse transfer neural network was input and a model for automatic customer classification and recognition was obtained. The results show that the as-proposed model can effectively achieve customer classification and sales evaluation,and its average detection accuracy can reach more than 95%.

关 键 词：云模型 云测度 离散余弦变换 价格敏感度 投诉率 欠费率 销售变现天数 忠诚度 决策支持 

分 类 号：TM76[电气工程—电力系统及自动化] TP393[自动化与计算机技术—计算机应用技术]