融合贷款信息和轨迹特征的催收评分卡模型  

Collection Scorecard Model Integrating Loan Information and Trajectory Features

作　　者：王延松 张华伟 张磊 王蒙 陈剑 WANG Yansong;ZHANG Huawei;ZHANG Lei;WANG Meng;CHEN Jian(Chery HuiYin Motor Finance Service Co.,Ltd.,Wuhu 241000;Yangtze River Delta Information Intelligence Innovation Research Institute,Wuhu 241000)

机构地区：[1]奇瑞徽银汽车金融股份有限公司,芜湖241000 [2]长三角信息智能创新研究院,芜湖241000

出　　处：《计算机与数字工程》2024年第10期2977-2983,共7页Computer & Digital Engineering

基　　金：2021年安徽省重点研究与开发计划(编号:202104a05020071)资助。

摘　　要：针对现有汽车金融中车辆贷后逾期风险识别准确率低、效率低等问题,论文提出了一种XGBoost-SMOTE的逾期风险预测模型。该模型通过采用SMOTE的过采样方法构建相对均衡的不同类别比例数据集,提高了模型的泛化能力。同时,结合了车辆用户基本信息、贷款信息和车辆GPS轨迹信息来进行特征工程,将融合后的特征数据输入到XGBoost分类器中,最终完成了车辆贷后逾期风险识别,实验结果表明:相较于其他机器学习算法预测模型,该模型可以达到更好的预测性能。This paper proposes a post-loan overdue risk of vehicle prediction model based on XGBoost-SMOTE to address the problems of the existing monitoring methods which can not identify risks effectively.In this paper,over sampling method of SMOTE is used to build a relatively balanced positive and negative proportion data set,which improves the generalization ability of the mod-el.And combined with the basic information of vehicle users,loan information and vehicle GPS track information for feature engi-neering.Then,the integrated feature data is input into XGBoost classifier,and finally complete the identification of post-loan over-due risk of vehicle.Compared with other machine learning algorithm prediction models,this model can achieve better prediction per-formance after the experiment.

关 键 词：汽车金融 贷后逾期 GPS轨迹 风险特征知识 

分 类 号：F832.4[经济管理—金融学] TP391[自动化与计算机技术—计算机应用技术]