随着农村产业结构的变化，越来越多的农户开始向金融机构申请贷款， 以发展规模经营，以实现收入增长。作为服务三农的主力军，甘肃农信适时 推出农户小额信用贷款业务，但由于目前农村金融体系尚不完善，以及农户 信用风险的特殊性，信贷业务逾期率居高不下，导致部分网点信贷资产严重 缩水，影响了金融机构对农户小额信用贷款的正常发放。甘肃农信农户小额 信贷业务包括评级授信、贷款发放、贷后管理三个阶段，以农户评级授信结 果为依据来评价其还款意愿和还款能力，农户信用等级评定是否准确，对业 务风险管理效果具有很大影响。 本论文依据风险管理研究框架，利用数据挖掘工具识别甘肃农信农户小 额信贷业务中的关键风险点，对之进行量化评估，并制定相应的风险管理策 略，以促进业务的持续健康发展。首先，论文选取甘肃农信 2016 年 8 月至 2017 年 8 月期间的 272132 条农户小额信贷样本数据（其中种植业信贷样本数据 151419 条，养殖业信贷样本数据 120713 条），以信用贷款是否逾期为目标变 量，以目前授信评级体系中的 18 个指标为风险特征变量，借助 SPSS19.0 统 计分析软件、选用 Logistic 回归进行分析，对种植业、养殖业用途的两个信贷 子类业务分别进行风险点识别。其次，结合现实数据，对识别出的风险点从 逾期概率及损失规模两个维度进行量化评估，进而进行风险分类。最后，根 据量化结果提出针对性的风险管理策略。 数据挖掘结果表明，种植业信贷业务有 17 个导致逾期的关键风险点，养 殖业信贷业务有 12 个导致逾期的关键风险点，两个子类信贷业务的风险类型、 风险表现及风险结构截然不同，差异非常显著。这意味着，当前农户小额信 贷“一揽子”的管理模式存在明显弊端，无论授信评级还是贷后监管，进行 风险分类管理更为合理；分类管理有助于提高风险识别的准确性和有效性， 在此基础上设计相应的风险管理策略将具有更好的针对性和有效性。 关键词：甘肃农信，农户小额信贷，风险量化，Logistic 回归兰州大学硕士学位论文 甘肃农信农户小额信贷业务逾期风险识别与控制策略研究 II OVERDUE RISK IDENTIFICATIONAND CONTROL STRATEGIES OF PEASANT HOUSEHOLD MICRO-CREDIT IN GANSU RURALCREDIT UNION Abstract With the change of rural industrial structure, more and more farmers begin to apply to financial institutions for loans to develop scale operation in order to achieve income growth. As the main force serving agriculture, rural areas and farmers, Gansu Rural Credit Union timely launched peasant household micro-credit. However, due to the imperfection of the rural financial system and the particularity of farmers' credit risk, the overdue rate of credit business remains high, resulting in a serious shrinkage of credit assets in some outlets, which affects the normal issuance of micro-credit loans to farmers by financial institutions. Gansu Rural Credit Union Peasant household micro-credit includes three stages: credit rating, loan issuance and post-loan management. Based on the results of credit rating, farmers' willingness to repay and repayment ability are evaluated. Whether peasant household micro-credit rating is accurate or not has a great impact on the effect of business risk management. According to the research framework of risk management, this paper uses data mining tools to identify the key risk points in the micro-credit business of rural credit farmers in Gansu Province, quantitatively evaluate them, and formulate corresponding risk management strategies to promote the sustainable and healthy development of business. Firstly, the paper selected 272,132 sample data of peasant household micro-credit from August 2016 to August 2017 of Gansu Rural Credit Union (including 151,419 sample data of planting credit and 120,713 sample data of farming credit). The target variable was whether the credit loan was overdue or not. The 18 indicators in the current credit rating system were taken as risk characteristic variables, and the SPSS19.0 statistical analysis software was used to select the data. Logistic regression analysis was used to identify the risk points of the two credit sub-businesses used in farming and aquaculture. Secondly, combined with the actual data, the identified risk points are quantitatively evaluated from the two dimensions of overdue probability and loss scale, and then risk classification is carried out. Finally, according to the quantitative results, we put forward targeted risk management strategies.兰州大学硕士学位论文 甘肃农信农户小额信贷业务逾期风险识别与控制策略研究 III The results of data mining show that there are 17 key overdue risk points in the farming credit business and 12 key overdue risk points in the farming credit business. The risk types, risk performance and risk structure of the two sub-types of credit business are quite different, and the difference is very significant. This means that there are obvious drawbacks in the current management model of package of micro-credit for farmers. Regardless of credit rating or post-loan supervision, risk classification management is more reasonable. Classification management helps to improve the accuracy and effectiveness of risk identification. On this basis, the design of corresponding risk management strategies will have better pertinence and effectiveness. Key words: Gansu Rural Credit Union, Farmer's Microfinance, Risk Quantification, Logistic Regression兰州大学硕士学位论文 甘肃农信农户小额信贷业务逾期风险识别与控制策略研究 IV 目 录 中文摘要...............................................I Abstract .............................................. II 第一章 引 言 ..........................................1 1.1 选题背景和研究意义 .......................................1 1.2 研究内容与思路 ...........................................3 第二章 相关理论概述 ....................................6 2.1 小额信用贷款 .............................................6 2.2 农村金融发展 .............................................8 2.3 金融风险管理 .............................................9 2.4 风险控制理论 ............................................10 第三章 甘肃农信农户小额信贷风险管理现状及问题分析 .....13 3.1 农户小额信贷业务流程简介 ................................13 3.2 农户小额信贷业务风险管理现状 ............................14 3.3 农户小额信贷业务风险管理问题 ............................19 第四章 甘肃农信农户小额信贷业务逾期风险分析 ...........21 4.1 业务分类及样本数据 ......................................21 4.2 模型选择 ................................................27 4.3 种植业农户小额信贷业务逾期风险分析 ......................30 4.4 养殖业农户小额信贷业务逾期风险分析 ......................43 第五章 甘肃农信农户小额信贷逾期风险管理策略与实施 .....53 5.1 风险管理策略的基本依据 ..................................53 5.2 种植业类信贷业务逾期风险点解释和管理策略 ................54兰州大学硕士学位论文 甘肃农信农户小额信贷业务逾期风险识别与控制策略研究 V 5.3 养殖业类信贷业务逾期风险点解释和管理策略 ................59 5.4 风险管理策略实施的保障措施 ..............................63 第六章 结论和展望 .....................................67 6.1 主要结论 ................................................67 6.2 研究展望 ................................................68