基于拒绝推论的小企业信用评分模型研究

在小企业信用评分模型的构建中,因数据缺失和样本选择性偏差可能导致模型参数估计有偏,对模型的预测能力和应用会有很大影响。本文利用从万德数据库中筛选出的小企业信息资料,模拟银行信贷筛选,产生带有缺失数据的模拟信贷样本,利用Heckman二阶段模型预测新的信用评分模型,将其结果与忽略缺失数据的审查模型和基于完全信息的标准模型进行比较。结果显示,Heckman二阶段模型的表现优于直接忽略缺失样本数据的审查模型,更接近标准模型的结果。这表明拒绝推论能够有效解决信用评分建模中数据缺失导致的样本选择偏差,提高信用评分模型的有效性和预测能力。

Research on Small Business Credit Scoring Model

Sample selection bias due to defects of sample data in the process of constructing small business credit scoring model may lead to biased parameter estimation,thereby affects the accuracy of model prediction and the effectiveness of credit decisions.Therefore,to improve the sample selection bias is the crucial content of current credit scoring model studies.Based on the statistics data of 171 small businesses from Wind financial database,which is divided into two subsamples randomly,we build a credit scorin...