基于风险等级的中小微企业信贷模型研究
2021-08-05顾一凡黄莉媛林晨欣曹春萍
顾一凡 黄莉媛 林晨欣 曹春萍



摘 要:为切实解决中小微企业贷款融资和银行对中小微企业贷款策略之间存在的问题,提出了基于风险等级的中小微企业信贷模型。该模型创新性地将机器学习算法引入传统中小微企业信贷风险及策略的研究当中,运用PCA降维、K-means聚类确定企业风险等级;通过Fisher线性判别确定银行信贷利率。应用该模型将123 家中小微企业分成五类风险等级,并给出银行对五类不同风险等级企业的贷款额度及利率,并通过实验验证模型的有效性和正确性。
关键词:K-means聚类;PCA降维;Fisher线性判别;信贷模型
中图分类号:TP391 文献标识码:A
Research on the Credit Model of Small, Medium and
Micro Enterprises based on Risk Level
GU Yifan1, HUANG Liyuan2, LIN Chenxin2, CAO Chunping1
(1.School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China;
2.Business School, University of Shanghai for Science and Technology, Shanghai 200093, China)
guyifan2020@126.com; 948384993@qq.com; 1246506991@qq.com; 2213893844@qq.com
Abstract: In order to effectively solve the problems between loan financing of small, medium and micro enterprises and the bank's loan strategy for them, this paper proposes to build a credit model for small, medium and micro enterprises based on risk level. This model innovatively introduces machine learning algorithms into the research on credit risks and strategies of traditional small, medium and micro enterprises. PCA (Principal Components Analysis) dimensionality reduction and K-means clustering are used to determine enterprise risk level. Bank credit interest rate is determined by Fisher linear discriminant. Based on this model, 123 small, medium and micro enterprises are divided into five risk levels, and the bank's loan lines and interest rates for each level are given. Validity and accuracy of the model are verified through experiments.
Keywords: K-means clustering; PCA dimensionality reduction; Fisher linear discriminant; credit model
1 引言(Introduction)
中小微企业作为我国宏观经济的“毛细血管”,贡献了我国80%的就业岗位和60%的GDP[1-2]。但中小微企业自身资金实力弱,若想发展就需要获得银行提供的贷款支持,而银行放贷首要考虑的是贷款资金的安全,因此银行放贷资金安全与中小微企业贷款需求之间的矛盾便成为一个亟待解决的问题。
目前,银行为了解决这一问题,采取了如下策略:对中小微企业进行风险评估,对于风险等级在一定标准下的企业,给予放贷。在郑建华等[3]提出的研究企业信用评级模型中,使用了层次分析法进行评级模型的构建。在郝晓露等[4]提出的商业银行贷款的研究中,使用了灰色预测模型对贷款利率进行了预测。王薛[5]使用了AHP——模糊综合评价农村信用社农户贷款风险模型。仔细分析上述模型可以发现,每个模型都涉及大量计算和推断。……
