银行贷款对中小企业信用优化评估建模仿真

Modeling and Simulation of Credit Optimization Evaluation of Small and Medium-sized Enterprises by Bank Loans

研究解决中小企业信贷存在信息不对称而导致的融资难、不良率高的问题,建立高精度分类预测模型迫在眉睫,但是由于中小企业信用评估特征多而复杂且冗余度高,使得寻找评估分类器的最优评估特征集成为难点问题,也是导致目前评估分类精度不高的关键因素。为了解决以上难点问题得到高精度信用评估模型,在使用粒子群神经算法(PSO-BP)的前提下,提出使用专家领域知识产生的综合权重对原始数据集进行特征选取,以及提出基于L2范数改进的logistic回归分类器对原始数据集进行特征提取,将上述两者结果合并构成最优特征集导入PSO-BP学习,建立中小企业信用优化评估模型。实验结果表明,提出的优化评估模型提高了特征集质量,提高了信用评估的精度,从而验证了优化模型的有效性。

In order to get high - precision classification model, we propose that the synthetic weights generated by expert domain knowledge are used in feature selection of the original data set, and the logistic regression classifier based on improved L2 norm is proposed to extract the feature from the original data set based on PSO - BP. The a- bove two results are combined to form the optimal feature set for PSO - BP learning, and to establish the SMEs credit optimization evaluation model. The experimental results show that the proposed optimization model can improve the quality of feature set and the accuracy of credit evaluation, and verify the effectiveness of the optimization model.