面向高维和不平衡数据的供应链金融信用评价

针对供应链金融模式下中小企业的信用风险控制问题，提出了一种面向高维和不平衡数据的信用风险预测模型。首先，基于Pearson-XGBoost两阶段特征选择建立供应链金融信用评价指标体系；其次，通过改进的NM-SMOTE算法对数据集进行平衡化；最后，利用Focal loss函数对XGBoost算法改进，并通过改进的粒子群算法进行优化，从而建立最终的信用评价模型。通过实验结果表明，提出的INS-IPSO-FLXGBoost模型对于中小企业具有更好的预测效果，可以更有效地识别风险企业。

Credit evaluation of supply chain finance for high dimensional and unbalanced data

Aiming at the credit risk control of small and medium-sized enterprises in the supply chain finance model, this paper proposed a credit risk prediction model for high dimensional and unbalanced data. Firstly, based on the Pearson-XGBoost two-stage feature selection, this model established the supply chain financial credit evaluation index system. Secondly, with the help of the improved NM-SMOTE algorithm, it made the dataset balanced. Finally, it used the Focal loss function to improve the XGBoost algorithm, and optimized it by the improved particle swarm algorithm, thus established the final credit evaluation model. The experimental results show that the INS-IPSO-FLXGBoost model proposed in this paper has a better prediction effect for small and medium-sized enterprises, and can identify risky enterprises more effectively.