高维单调转移模型的变量选择及其在违约风险评估中的应用

项目名称： 高维单调转移模型的变量选择及其在违约风险评估中的应用

项目编号： No.71501089

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 管理科学

项目作者： 史兴杰

作者单位： 南京财经大学

项目金额： 15.5万元

中文摘要： 近年来，互联网金融的发展极大地促进了金融创新和普惠金融体系的建立，然而信用风险依然是互联网金融企业面临的重要风险。对于互联网金融中的高维风险因素，传统的违约风险评估模型并不适用。建立既可以挑选重要解释变量又能避免模型误判的统计模型进行违约风险评估是十分具有挑战性的问题。本项目旨在高维数据下，研究常系数单调转移模型、变系数单调转移模型、部分线性单调转移模型的变量选择问题。在不同的单调转移模型下，研究变量选择的统计性质，开发惩罚秩估计的求解算法，结合风险因素间的交互作用构建违约风险评估模型。这些方法是在高维的风险因素下进行违约风险评估的重要方法和技术，也是进行信用风险管理、控制的重要理论和方法基础。同时，建立有效的违约风险评估模型对预警风险、规避信用危机，乃至对金融安全均具有重要意义。

中文关键词： 统计学；变量选择；单调转移模型；违约风险评估；风险管理

英文摘要： Recently, the development of internet finance has greatly contributed to the financial innovation and the establishment of inclusive financial systems. However, the credit risk remains a significant risk faced by internet finance. For high-dimensional risk factors in internet finance, existing approaches are not suitable for default risk evaluation. It is a very challenge problem for default risk evaluation to build a robust statistic model, which can select the important determinants of defaults. For high dimensional data, this project focuses on variable selection in the monotone transformation model with constant coefficients, the monotone transformation model with varying coefficients, and the partial linear monotone transformation model. We study the variable selection property of proposed methods, develop efficient algorithms for penalized rank estimators, and construct a default risk evaluation model taking interactions between risk factors into account. The proposed research will not only develop important methods and techniques for the default risk evaluation with high-dimensional risk factors, but also enrich risk management and credit risk management theory and methodology. Therefore, the proposed research is extremely important to build efficient risk evaluation models, which can be used to provide early warning for default risks in order to avoid credit crisis and ensure financial security.

英文关键词： Statistics;Variable selection;Monotone transformation model ;Evaluation of default risk;Risk management