摘要
现实中,由于隐私保护的限制,使得对重要客户的识别十分困难.具体地,很难获取商业银行重要基金客户的具体标签信息,这给建立相关的预测模型带来了极大的挑战.然而,通过特定的客户群估计重要基金客户所占比例是可行的.因此,提出了一种基于比例标签学习的商业银行重要基金客户挖掘新方法.方法的特点在于仅仅使用样本标签比例信息(label proportions information)去构建分类模型,进而有效地识别商业银行中的重要基金客户.同时,大量的实验结果表明了该方法的有效性,这对于有效解决隐私保护下的重要基金客户识别问题提供了一种新途径,具有明显的现实意义及实践价值.
Due to the privacy restrictions in reality, it is very difficult to identify the important clients. Specifically, we cannot get the specific label of commercial banks＇ fund client, which indicates whether each individual is an important client or not. It brought great challenges for establishing related forecasting model. However, we can estimate the proportion of the important fund clients through specific clients group. Therefore, this paper proposed a new method based on proportion label learning for commercial banks to recognize important fund clients. This method can only use the proportion information of sample labels to learn a classification model, which would identify the important fund clients efficiently. Meanwhile,sufficient experimental results show the effectiveness of this method, which provided a new view to efficiently solve the problem of identifying important fund clients under privacy protection.It has distinct practical significance and value.
作者
石勇
马福海
齐志泉
崔荔蒙
SHI Yong MA Fu-hai QI Zhi-quan CUI Li-men(Research Center on Fictitious Economy ＆ Data Science, Chinese Academy of Sciences, Beijing 100190, China School of Management, Chinese Academy of Sciences, Beijing 100190, China Key Laboratory of Big Data Mining and Knowledge Management,Chinese Academy of Sciences, Beijing 100190, China)
出处
《数学的实践与认识》
北大核心
2017年第19期291-302,共12页
Mathematics in Practice and Theory
基金
国家自然科学基金项目(71331005
71110107026
91546201)
关键词
比例标签学习
模式识别
机器学习
重要客户识别
learning with label proportions
Pattern Recognition
Machine Learning
Theimportant client identification
