个人信用评估方法优化及其在我国应用研究

发布时间：2018-12-08 16:20

【摘要】：个人信用评估问题是风险管理的重要组成部分,随着社会的发展,个人信用评估问题已经得到了广泛的重视和应用。它可以被转化成一个两分类问题,解决两分类问题常用的方法有很多,例如：回归算法、近邻算法、遗传算法、神经网络和支持向量机等。其中,支持向量机已经成为个人信用评估方法的研究热点。本文对支持向量机进行推广,得到了处理矩阵数据分类的支持矩阵机方法。为了能解决非线性分类问题,本文又引用了矩阵数据核函数,然后将矩阵数据核函数和支持矩阵机进行组合,就得到了处理个人信用评估的新方法。本文的研究内容及结果：(1)简单叙述了分类问题以及支持向量机理论。(2)传统的个人信用评估数据都是向量型数据,而这种向量型数据有两大不足：一是丢失指标之间的关系信息；二是不具有时间动态性,这两点不足都会极大的影响最终的信用评估结果。(3)针对向量型数据的不足,我们提出了个人信用评估的矩阵型数据,并且给出了构造矩阵型数据的两种方法,当然这两种方法也不是完满的,我们还对这两种方法做了分析和比较,给出了各自的优缺点和适用的信用评估问题。(4)对我们提出的加核支持矩阵机方法,在国外信用评估数据集German、 Australian和国内住房贷款信用数据集上进行试验,结果证明我们的优化方法的有效性。
[Abstract]:Personal credit assessment is an important part of risk management. With the development of society, personal credit evaluation has been widely paid attention to and applied. It can be transformed into a two-classification problem. There are many common methods to solve the two-classification problem, such as regression algorithm, nearest neighbor algorithm, genetic algorithm, neural network and support vector machine. Among them, support vector machine (SVM) has become the focus of personal credit evaluation. In this paper, the support vector machine (SVM) is generalized and the support matrix machine (SVM) method is proposed to deal with the classification of matrix data. In order to solve the nonlinear classification problem, a new method of dealing with personal credit evaluation is obtained by combining the matrix data kernel function with the support matrix machine. The main contents and results of this paper are as follows: (1) the classification problem and the support vector machine theory are briefly described. (2) the traditional personal credit evaluation data are vector data. But this kind of vector type data has two big shortcomings: one is to lose the relation information between the index; Second, there is no time dynamic, and these two shortcomings will greatly affect the final credit evaluation results. (3) aiming at the shortage of vector data, we propose the matrix data of personal credit evaluation. And two methods of constructing matrix data are given. Of course, these two methods are not perfect, and we also analyze and compare the two methods. The advantages and disadvantages of each method and the applicable credit evaluation problems are given. (4) We test the credit evaluation data set (German, Australian) and the domestic housing loan credit data set (German, Australian), which are proposed by us. The results show that our optimization method is effective.
【学位授予单位】：首都经济贸易大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.4

【相似文献】