Prediction of Loan Rate for Mortgage Data: Deep Learning Versus Robust Regression
Prediction of Loan Rate for Mortgage Data: Deep Learning Versus Robust Regression
复制标题
DOI：
10.1007/s10614-022-10239-5
复制
发表时间：
2022-02
影响因子：
2
通讯作者：
Donglin Wang;Don Hong;Qiang Wu
中科院分区：
文献类型：
--
作者：
Donglin Wang;Don Hong;Qiang Wu
抵押贷款数据往往是扭曲的，有缺失的信息，并受到离群值的污染。当抵押贷款公司或银行对新申请人的票据利率进行预测时，通常会选择稳健的回归模型来处理异常值。在本文中，我们利用深度神经网络来预测贷款利率，并将其性能与三个经典的鲁棒回归模型进行比较。两个真实的抵押贷款数据集用于此比较。结果表明，深度神经网络具有最佳性能，因此推荐使用。
Mortgage data is often skewed, has missing information, and is contaminated by outliers. When mortgage companies or banks make prediction of note rates for new applicants, robust regression models are usually selected to deal with outliers. In this paper, we utilize deep neural network to predict the loan rate and compare its performance with three classical robust regression models. Two real mortgage data sets are used in this comparison. The results show that deep neural network has the best performance and therefore is recommended.