基于KLIEP-Logistic的中小微企业信用评价研究——以浙江省模具行业小样本企业为实证

Research on Credit Evaluation of SMEs with Few-Shot Situations Based on KLIEP-Logistic Algorithm——Taking the Enterprises With Few-Shot Situations in the Mold Industry in Zhejiang Province as Evidence

中小微企业在细分行业下广泛存在样本量低、正负样本不均衡等小样本问题,直接应用传统的数据驱动信用评价模型往往难以达到理想效果。本研究根据相似行业的企业具有数据分布相近的特性,融合迁移学习理论,提出一种基于KLIEP-Logistic的小样本中小微企业信用评价方法。通过对浙江省A市制造业行业下存在小样本问题的模具细分行业进行实证实验,结果表明:与传统数据模型对比,基于KLIEP-Logistic构建的信用评价模型对小样本中小微企业客群的信用预测效果更好,具有更强的预测能力和泛化能力。

SMEs(Small, medium and micro-enterprises) are widely faced with few-shot situations such as small sample size and imbalanced data in subdivided industries. It is often difficult to achieve ideal results by directly applying traditional data-driven credit evaluation models. According to the characteristics of similar data distribution of enterprises in similar industries, this research integrates the transfer learning theory, and proposes a credit evaluation method based on KLIEP-Logistic algorithm for few-shot situations of SMEs. Through empirical experiments on the mold subdivision industry with fewshot situations in the manufacturing industry in Zhejiang Province A city, the experimental results show that, compared with the traditional data-driven model, the credit evaluation model constructed based on KLIEP-Logistic algorithm has better prediction on the customer group of SMEs, and it has stronger accuracy and generalization ability.