主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

Chinese Journal of Management Science ›› 2022, Vol. 30 ›› Issue (3): 106-116.doi: 10.16381/j.cnki.issn1003-207x.2021.1324

• Articles • Previous Articles     Next Articles

Research on the Customer Value Portrait Model of Industrial Power Enterprise in China Based on Spectral Clustering Technology and Rough Set Theory

YU Shun-kun1, YAN Hong-xu1, DU Shi-yue2, LIN Yi-qing3   

  1. 1. School of Economics and Management, North China Electric Power University, Beijing 102206, China;2. Pumped-storage Technological & Economic Research Institute, State Grid Xinyuan Co., Ltd., Beijing 100761, China; 3. Shantou Power Supply Bureau of Guangdong Power Grid Co., Ltd., Shantou 515000, China
  • Received:2021-07-06 Revised:2021-08-05 Online:2022-03-19 Published:2022-03-19
  • Contact: 闫泓序 E-mail:hxyan@ncepu.edu.cn

Abstract: For Chinese power supply enterprises, the value portrait of power customers has important practical significance for improving the resource allocation efficiency of marketing service, promoting the smart marketing management, and therefore maximizing the comprehensive benefits. However, the reported models fail to reflect the development requirements of China’s industrial enterprises in the latest national energy policy, cannot cope well with the sparsity of real power customers’ electricity consumption data, and has room for improvement in the accuracy of new customers’ value rating prediction. Therefore, based on the spectral clustering (SC) technology and the rough set (RS) theory, an optimized industrial power customer value portrait model termed as SC-RS is constructed. The new model is constructed according to the logical framework of “knowledge extraction-knowledge reasoning-knowledge service”. Firstly, in the knowledge extraction part, an optimized China’s industrial power customer rating index system is constructed, based on the China’s “Peak Carbon Dioxide Emissions” goal and “Carbon Neutrality” vision. In addition, the grid search strategy combined with the SC technology is used to refine user’s value grade information. Secondly, in the knowledge reasoning part, the RS theory is used to build a four-dimensional rule mining model to generate the rule base of the user value grade, based on the framework of three-dimensional rule mining. It consists of row reduction based on the ChiMerge discretization method and coefficient of variation, column reduction based on the system dependence, cell reduction based on the object certainty factor, and rule extraction based on the rule strength. Finally, in the knowledge service part, the user value grade intelligence is applied to portray the value portrait of group users. The rule base is used to present understandable value knowledge and construct a ruled-soft classifier to achieve value grade prediction and individual value portrait description of new users. The model is further applied on the data of actual industrial power customers. The results show that the constructed rating index system closely copes with the latest developments in Chinese power industry. The SC-RS model is compatible with sparse data and has low data requirements. The constructed four-dimensional rule mining model based on RS theory is feasible and effective, and can robustly predict the value grade of new customers. Moreover, the SC-RS model can realize the value intelligence mining and utilization of big data of electric power customers, which is a powerful tool for empowering the smart marketing management of China’s electric power enterprises.

Key words: smart marketing management, power customer value portrait model, rough rule mining model, value grade prediction, carbon peak, carbon neutral

CLC Number: