基于流形学习的企业信用风险组合评价模型

Combined Evaluation Model of Enterprise Credit Risk Based on Manifold Learning

为精准预测企业潜在的信用风险,构建基于流形学习的信用风险评价模型。首先,计算企业违约情况与财务指标的相关系数,剔除掉相关性弱的指标。其次,基于流形学习的局部线性嵌入方法对剩余指标数据进行约简,利用贝叶斯模型、决策树模型和BP神经网络模型对企业的信用风险进行分类评价,构建基于诱导有序集成的组合评价模型。对300家创业板上市企业数据进行仿真分析,为验证模型的有效性,在300家公司中(其中270家为训练样本,30家为测试样本)随机选取2组样本,使用ST公司被执行特别处理(special treatment,ST)前一年的数据进行测试,结果表明组合模型具有更高的稳定性和分类精度。

Constructing a scientific and reasonable credit risk evaluation model to accurately predict the potential credit risks of enterprises can provide technical support for enterprises to formulate relevant policies and reduce investment risks for investors.Therefore,this paper firsty calculates the correlation coefficient between corporate defaults and financial indicators,and removes indicators with weak correlations.Secondy,the residual index data is reduced by a local linear embedding method based on manifold learning.Then the Bayesian model,decision tree model and BP neural network model are used to classify and evaluate the credit risk of enterprises.Considering that the evaluation accuracy of each single evaluation method for different enterprises is different,in order to comprehensively utilize the advantages of each method,a combined evaluation model based on induced ordered integration is constructed.Finally,the simulation analysis is carried out on the data of 300 companies listed on the Growth Enterprise Market.In order to verify the validity of the model,2 groups of samples are selected from the 300 companies(270 of which are training samples and 30 are testing samples)randomly,using ST company.The data from one year before the special treatment(ST)is performed are tested,and the experimental results show that the combined model has higher stability and classification accuracy.