银行的企业贷款违约风险预测分析

发布时间：2018-04-04 07:53

本文选题：企业违约风险分析　切入点：隐私保护　出处：《电子科技大学》2012年硕士论文

【摘要】：如何在保护数据私密性的前提下,更有效地发现具有潜在违约风险的企业贷款是银行业和学术界的热门研究方向。本文针对企业贷款数据分布不平衡的特征,提出了一套在保护隐私前提下的综合使用多目标决策方法选择最优分类预测模型的方法。本次研究所提出的方法在预处理部分主要是为了使得整个研究方法在实际操作中的原始数据更加干净,剔除因为数据原因造成的不良影响。进行维度规约则主要是为了防止进行数据分析操作时,一些敏感核心数据可能会泄露,从而对银行和客户造成重大损失。整个方法的数据分析部分则采取不平衡数据的过采样、分类算法选择、分类评价指标选择、多目标决策方法选择等步骤,构建一个合适的分类预测模型,用以帮助银行提高分辨贷款违约企业的准确率,降低银行的贷款风险,提高银行的收益。我们以中国某国有银行四川省分行的企业贷款数据为例,以PCA作为隐私保护的一种方法。同时我们又得到了在加权情况下,最优多目标决策方法TOPSIS的排序要比没有进行分类评估指标加权来的更加真实可信。
[Abstract]:On the premise of protecting the privacy of data, it is a hot research direction of banking and academic circles to find the enterprise loan with potential default risk more effectively.In view of the unbalanced distribution of enterprise loan data, this paper proposes a comprehensive multi-objective decision making method to select the optimal classification and prediction model under the premise of privacy protection.The main purpose of this research is to make the raw data of the whole research method cleaner in practice and eliminate the bad effects caused by the data.The main purpose of dimension specification is to prevent some sensitive core data from leaking, which will cause great losses to banks and customers.In the data analysis part of the whole method, a suitable classification and prediction model is constructed by taking steps such as over-sampling of unbalanced data, selection of classification algorithm, selection of classification evaluation index, selection of multi-objective decision method, etc.In order to help banks to improve the accuracy of the resolution of loan default enterprises, reduce the bank's loan risk, improve the bank's income.We take the enterprise loan data from Sichuan branch of a state-owned bank in China as an example and use PCA as a privacy protection method.At the same time, we get that the ranking of the optimal multi-objective decision method TOPSIS is more real and credible than that without the weighted classification evaluation index.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.4;F224

【参考文献】