Abstract: How to efficiently recover consumers’ electric fees is a key task in power marketing. However, constructing appropriate payment channels may help different consumers fulfill their payments better. In this study, a novel method is proposed to categorize consumers’ payment preference based on black hole entropy fuzzy clustering and Bayesian model. The proposed method is very effective for massive user data, and can adaptively discovery the optimal preference clusters by means of maximum-a-posteriori reasoning and iterative sampling optimization. The experimental results obtained by using the proposed method on true sampled consumers’ payment records in Jiangsu province clearly demonstrate that, constructing more self-service payment channel is the most worthy, and the prepaid payment channel has most potential values.

Key words: electricity recovery, payment channel, black hole entropy, fuzzy clustering

摘要： 电费回收直接关系到电力公司的有序运行，而交费渠道的建设和用电客户交费渠道的偏好，直接影响电力公司的电费回收;因此做好用电客户交费渠道偏好的研究至关重要。基于黑洞熵理论和贝叶斯模型，提出了适合电力数据的黑洞熵模糊聚类模型。该模型的显著特点是能通过最大后验概率推理和迭代抽样算法求解聚类最优参数，且适用于大规模用户数据。基于黑洞熵模糊聚类对某省用电客户的实际抽样数据进行模式发现表明，电力公司自营自助交费渠道尤为值得推广，预付费交费渠道最具有潜力。

关键词: 电费回收, 交费渠道, 黑洞熵, 模糊聚类