针对信用评分的特点,首先对传统的Bagging算法进行改进,然后基于不同的基分类器构建多个改进的Bagging组合模型,最后根据一种全新的方式将其组合,得到双层组合模型.通过实证分析得出以下结论:一、与原Bagging算法相比,改进算法能有效降低第二类错误率;二、与单层组合模型相比,多层组合模型能进一步降低第二类错误率和损失率. Aiming at the characteristics of credit score, firstly, the traditional Bagging algorithm is improved, then a number of improved Bagging combination models are constructed based on different base classifiers, and finally a new combination of them is obtained according to a new way to obtain a double-layer combination model. The conclusions are as follows: 1. Compared with the original Bagging algorithm, the improved algorithm can effectively reduce the error rate of the second type. Second, compared with the single-layer combination model, the multi-layer combination model can further reduce the second error rate and loss rate.