基于自适应惯性权重PSO-LightGBM的信用风险评估研究  

Credit risk assessment based on LightGBM and adaptive inertia weight PSO

作　　者：付芷宁 李慧敏[1] 徐亚田 陶玉虎 高伟 FU Zhi-ning;LI Hui-min;XU Ya-tian;TAO Yu-hu;GAO Wei(School of Mathematics and Computer Science,Yunnan Minzu University,Kunmin,China)

机构地区：[1]云南民族大学数学与计算机科学学院,云南昆明650500

出　　处：《云南民族大学学报（自然科学版）》2024年第3期345-350,共6页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：云南省研究生优质课程建设项目(云学位[2022]8号).

摘　　要：贷款市场复杂的个人信用风险问题中,信用风险评估模型的构建是十分关键的一步.利用Lending Club数据集,进行信用风险评估模型的构建来预测客户的违约概率.首先进行数据处理,再通过合成少数类过采样技术(SMOTE)算法处理数据正负样本不平衡的问题,获得完备的信用贷款数据.其次采用轻量梯度提升机(LightGBM)模型进行训练,并使用自适应惯性权重的粒子群优化(PSO)算法得到LightGBM的最优参数.与多个主流算法进行对比,实验结果表明,构建的模型有更好的性能.In view of the complex personal credit risk problem in the loan market,the construction of credit risk as⁃sessment model is a very important step.Using the Lending Club dataset,the credit risk assessment model is con⁃structed to predict the default probability of customers.First,data processing is carried out,and then the problem of positive and negative sample imbalance is processed by the SMOTE(synthetic minority oversampling technique)al⁃gorithm to obtain complete credit loan data.Secondly,the LightGBM model is used for training,and the PSO(parti⁃cle swarm optimization)algorithm of adaptive inertia weights is used to obtain the optimal parameters of LightGBM.After comparison with multiple mainstream algorithms,experimental results show that the constructed model has bet⁃ter performance.

关 键 词：信用风险 不平衡数据 合成少数类过采样技术 LightGBM模型 粒子群优化算法 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] F832.4[自动化与计算机技术—计算机科学与技术]