«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

期数:

2011年03期

页码:

213-216

栏目:

应用开发研究

出版日期:

1900-01-01

文章信息/Info

Title:

Application of Support Vector Machines in Personal Credit Rating

文章编号:

1673-629X（2011）03-0213-04

作者:

叶小娇 李汪根 黄尧颖

安徽师范大学数学计算机科学学院

Author(s):

YE Xiao-jiao;  LI Wang-gen;  HUANG Yao-ying

College of Mathematics and Computer Science ,An hui Normal University

关键词:

信用评估; 支持向量机; 不平衡数据; 分类

Keywords:

credit rating ;  support vector machines ;  unbalanced data;  classification

分类号:

TP181

文献标志码:

A

摘要:

个人信用评估在银行信贷业务中有着举足轻重的作用。为了提高银行对个人信用评估的准确率，将支持向量机应用到个人信用评估中，以德国信贷数据为数据集，采用网格-5折交叉验证方法获取核函数最优参数，然后选择不同的核函数及其最优参数进行训练建模，实验得出RBF核函数更适合该数据集。针对样本中数据不平衡的问题，通过改变权重的方式对不同类别设置不同的惩罚参数。实验结果表明，该方法在保证总的预测准确率较好的前提下，有效地平衡了第一类和第二类错误率，可以作为银行信贷决策的参考依据

Abstract:

Personal credit rating plays a vital role in bank credit business. In order to increase personal credit rating accuracy, support vector machines （SVM） is used to solve the problem of personal credit rating prediction in this paper. With the data set about credit from Germany, optimal parameters are obtained using 5-cross validation via parallel grid search, then four different kernel functions are selected to train the date set. The experimental results demonstrate that RBF kernel function is more suitable for the data set. As the data set is unbalanced, the rates of the first class error and the second class error are efficiently balanced by setting different punishment for different datasets under the premise of better overall prediction accuracy. It can be used as reference for the bank credit decisions

备注/Memo

备注/Memo:

安徽省自然科学研究重点项目（KJ2010A140）叶小娇（1983-），女，浙江台州人，硕士研究生，研究方向为智能算法及其应用；李汪根，硕士生导师，研究方向为生物计算和智能计算

常用功能

更新日期/Last Update: 1900-01-01