欢迎访问中国科学院大学学报,今天是

中国科学院大学学报 ›› 2013, Vol. 30 ›› Issue (3): 298-303.DOI: 10.7523/j.issn.1002-1175.2013.03.003

• 数学 • 上一篇    下一篇

基于改进RBF神经网络的银行个人信用评级

蓝润荣, 程希骏   

  1. 中国科学技术大学统计与金融系, 合肥 230026
  • 收稿日期:2012-04-25 修回日期:2012-10-11 发布日期:2013-05-15
  • 通讯作者: 蓝润荣, lanrr@mail.ustc.edu.cn
  • 基金资助:

    中国科学院知识创新工程重要方向项目(KJCX3-SYW-S02)资助 

A new RBF neural network and its application on individual credit rating in banks

LAN Run-Rong, CHENG Xi-Jun   

  1. Department of Statistics and Finance, University of Science and Technology of China, Hefei 230026, China
  • Received:2012-04-25 Revised:2012-10-11 Published:2013-05-15

摘要:

研究RBF神经网络在个人信用评级中的应用.针对传统的RBF神经网络无法处理非数值型数据和对初始中心的选取及异常值十分敏感等问题,提出一种基于模糊K-Prototypes算法的RBF神经网络,提高了处理分类型数据及混合型数据的能力,并且改进的模糊K-Prototypes算法有助于降低模型对初始中心选取和异常值的敏感性.将改进前后的模型分别应用于商业银行的个人信贷评级中,结果表明,改进后的模型预测精度和稳健性都优于传统的RBF模型.

关键词: RBF神经网络, 模糊K-Prototypes算法, 分类型数据, 信用评级

Abstract:

We mainly focus on the application of RBF neural networks in individual credit evaluation. Considering that the traditional RBF neural networks can only deal with the numerical value and that it is sensitive to the noisy data and the initial clustering centers, we propose a new RBF neural network combined with fuzzy K-Prototypes algorithm, which can deal with mixed data and is less sensitive to the noisy data and the initial clustering centers. The experimental results on the credit data show that the new RBF neural network has higher accuracy and robustness than the traditional one.

Key words: RBF neural networks, fuzzy K-Prototypes algorithm, categorical data, credit rating

中图分类号: