基于LightGBM信贷风控模型的算法优化  

ALGORITHM OPTIMIZATION OF CREDIT RISK CONTROL MODEL BASED ON LIGHTGBM

作　　者：吴照明 胡西川[2] Wu Zhaoming;Hu Xichuan(School of Information Engineering,Shanghai Maritime University,Shanghai 201306,China;Department of Computer,School of Information Engineering,Shanghai Maritime University,Shanghai 201306,China)

机构地区：[1]上海海事大学信息工程学院,上海201306 [2]上海海事大学信息工程学院计算机系,上海201306

出　　处：《计算机应用与软件》2022年第6期342-349,共8页Computer Applications and Software

摘　　要：对于如今的金融机构如何迅速地对信贷风险做出准确判断并进行有效地控制是十分关键的。基于LightGBM算法建立信贷风控模型,对借贷人的个人信息进行数据的清洗筛选和特征衍生。融合pair-wise算法,可以优化特征的排序,防止过拟合。实验结果表明,相较于XGBoost分类算法,基于LightGBM算法建立的信贷风控模型的预测精度提高12.3%。该算法的信贷风控模型占用内存较少,支持并行处理,具有一定的参考价值。It is very important for today s financial institutions to quickly make accurate judgment and effective control of credit risk.Based on the LightGBM algorithm,the credit risk control model was established to clean and filter the personal information of the borrower and derive the characteristics.It could optimize the sorting of features and prevent over fitting by combining the pair-wise algorithm.The experimental results show that compared with XGBoost classification algorithm,the prediction accuracy of the proposed model is improved by 12.3%.The proposed model takes less memory and supports parallel processing,which has certain reference value.

关 键 词：互联网金融 风控模型 LightGBM 数据挖掘 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]