面向农户贷款信用风险评估的Relief-GEP模型  

A Relief-GEP Model for Credit Risk Evaluation of Loans for Farmers

在线阅读下载全文

作  者:刘田田 LIU Tian-tian(College of Information and Electromechanical Engineering,Jiangsu Open University,Nanjing 210017,China)

机构地区:[1]江苏开放大学信息工程学院,江苏南京210017

出  处:《软件导刊》2020年第11期79-83,共5页Software Guide

基  金:江苏省高等学校自然科学研究项目(18KJB520008);江苏开放大学“十三五”青年项目(16SSW-Q-003)。

摘  要:现有的信用评估模型往往需要在建模前预设模型的基本形态结构,极易因函数形式的误设导致评估结果误差较大,同时现有模型大多面向传统大中型企业、消费信贷用户,对农户这一借款主体缺乏适用性。基于宿迁地区某农村信用社的农户信贷数据库,设计一种面向我国农户贷款信用风险评估模型——Relief-GEP模型。模型首先使用Relief算法,对建模样本集维度进行删减,剔除对预测违约概率影响不大的各项指标,在摈弃部分噪声数据的同时提高后续建模精度;在缺乏最优函数形式的先验信息情况下使用基因表达式编程算法,以"适者生存"的逻辑通过反复演化迭代,动态构筑模型的核心部分。实证研究表明,Relief-GEP模型相比于当前流行的12个信用风险评估模型,拥有更优的拟合精度与更好的泛化能力。The existing credit evaluation model needs to presuppose the basic structure of the model itself before the modeling process,then the error of the evaluation result is very easy to be caused by the setting error of the function form of the model.At the same time,most of the existing models are designed for traditional large and medium-sized enterprises or consumer users,and are lack of applica⁃bility to farmers who wish to apply for loans.Based on the credit database of a rural credit cooperative in Suqian area,this paper de⁃signs a credit risk evaluation model-the Relief-GEP model which is oriented to evaluate the credit risk of loans for farmers in China.The model firstly uses the relief algorithm to reduce the dimension of the sample set for modeling,eliminate all the indexes that have lit⁃tle effect on the prediction of default probability and some noise data to improve the accuracy of subsequent modeling process,and then in the condition of lack of the prior information of the optimal function form,the gene expression programming algorithm is employed with the logic of“survival of the fittest”and dynamically construct the model through the repeated evolutionary iteration.The empirical study shows that the Relief-GEP model has better fitting precision and better generalization ability than the 12 credit risk evaluation models that are popular in the current academic and industry circles.

关 键 词:仿真建模 信用风险 特征权重选择算法 基因表达式编程算法 

分 类 号:TP303[自动化与计算机技术—计算机系统结构]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象