基于模糊信息分解的中小微企业信用评估组合模型

A Combination Model for Credit Evaluation of Small, Medium and Micro Enterprises Based on Fuzzy-based Information Decomposition

信用评估是银行开展中小微企业信贷业务的重要环节,尽管已有信用评估模型的分类性能较高,但仍有待进一步提升,同时模型也缺乏可解释性,在一定程度上影响了模型的推广应用。为此,本文提出了一种基于模糊信息分解的中小微企业信用评估组合模型。首先,针对信用评估数据的不平衡性,利用模糊信息分解的方法进行过采样处理。其次,基于XGBoost模型和Logistic回归模型,构建了一种具有较高的分类性能和良好可解释能力的中小微企业信用评估组合模型。最后,结合某商业银行对中小微企业的信用数据,实验验证了所构建组合模型的可行性和有效性。

Credit evaluation is an important link for banks to carry out credit business for small,medium and micro enterprises.Although the existing credit evaluation models have achieved a high performance,there are still some limitations,such as the performance needs to be further improved and the interpretability needs to be enhanced,which affect the applications of these models to some extent.Therefore,this paper proposes a combination model for credit evaluation of small,medium and micro enterprises based on fuzzy-based information decomposition.Firstly,in view of the imbalance of credit evaluation data,fuzzy-based information decomposition is used to conduct over-sampling.Secondly,combining Logistic model and XGBoost model,a combination model with high performance and good interpretability is constructed for the credit evaluation of small,medium and micro enterprises.Finally,experiments on the credit data of a commercial bank for small,medium and micro enterprises,verify the feasibility and effectiveness of the proposed model.