基于机器学习的农户农地经营权抵押贷款信用风险识别及其损失度量  

Identification and loss measurement of credit risk on rural households’farmland management right mortgages based on the machine learning

作　　者：彭艳玲[1] 彭一杰 周红利 汪寿阳[5,6,7] 蒋远胜 PENG Yanling;PENG Yijie;ZHOU Hongli;WANG Shouyang;JIANG Yuansheng(College of Economics and Management,Northwest A&F University,Yangling 712100,China;Guanghua School of Management,Peking University,Beijing 100871,China;Institute for Artificial Intelligence,Peking University,Beijing 100871,China;College of Economics,Sichuan University,Chengdu 610065,China;Academy of Mathematics and Systems Science,Chinese Academy of Sciences,Beijing 100190,China;Center for Forecasting Science,Chinese Academy of Sciences,Beijing 100190,China;School of Economics and Management,University of Chinese Academy of Sciences,Beijing 100190,China;College of Economics,Sichuan Agricultural University,Chengdu 611130,China)

机构地区：[1]西北农林科技大学经济管理学院,杨凌712100 [2]北京大学光华管理学院,北京100871 [3]北京大学人工智能研究院,北京100871 [4]四川大学经济学院,成都610065 [5]中国科学院数学与系统科学研究院,北京100190 [6]中国科学院预测科学研究中心,北京100190 [7]中国科学院大学经济与管理学院,北京100190 [8]四川农业大学经济学院,成都611130

出　　处：《系统工程理论与实践》2025年第2期448-462,共15页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金青年项目“农村承包土地经营权抵押贷款信用风险生成机理及分担机制研究”(71903141);国家自然科学基金杰出青年科学基金“智能管理系统仿真与优化”(72325007);国家自然科学基金优秀青年科学基金“管理中的仿真优化”(72022001);国家自然科学基金原创探索项目“智能系统的新学习方法”(72250065)。

摘　　要：在农户风险持续暴露背景下,立足中国土地产权管制与农村金融生态不完备情境,本文依据宁夏、重庆、四川三省份农地经营权抵押融资试点地区入户调查数据,运用机器学习方法识别农户信用风险,并验证了该方法与传统模型相比的有效性,在此基础上进一步采用CreditRisk+模型度量了农户信用风险损失.调查统计显示,农户农地经营权抵押贷款违约率较高,为10%,且以主动违约为主.研究发现,机器学习方法中的随机森林模型可有效识别信用风险关键因素和预测违约概率.进一步地,农户农地经营权抵押贷款信用风险损失和单笔贷款风险敞口较高,且在极端事件冲击下损失增幅明显.此外,现有风险控制框架下增加农户被动性违约动机的考察,有助于金融机构优化金融资本结构,完善风险控制策略.据此,提出加快金融科技发展、完善农村征信体系、创新风险预警工具等建议.Using the survey data collected from rural households in Ningxia,Chongqing,and Sichuan provinces,this paper has identified the credit risk and measured the risk loss,under the context of land property rights controlled and the imperfect ecology of rural finance market in China.This paper uses machine learning method to identify farmers’credit risk and verifies the effectiveness of this method compared with the traditional model.Also,Credit Risk+model is employed to evaluate farmers’credit risk.According to the survey statistics,the default rate of farmers’farmland management right mortgages is relatively high,and it was 10%.Results show that the random forest model could identify the key factors of credit risk and predict the default probability effectively.Moreover,the expected loss and risk exposure of each loan is relatively high,and the risk loss increases rapidly under the impact of extreme events.In addition,it is helpful for financial institutions to optimize the financial capital structure and improve the risk management strategy to increase the investigation of farmers’passive default motivation under the prior risk management framework.Thus,we conclude with several policy implications such as the accelerating development of fintech,improvement of rural credit investigation system,and innovation of risk pre-warning tools.

关 键 词：农地经营权抵押贷款 农户信用风险 风险识别 损失度量 机器学习 

分 类 号：F323.9[经济管理—产业经济]