Abstract: The cut-off point of default judgment is an important reference standard for whether financial institutions accept customer loan applications. An inappropriate cut-off point of default judgment may cause financial institutions to mistakenly accept a large number of loan applications from potential default customers, which in turn will cause huge losses and find a suitable one. The cut-off point of default judgment is of great significance to the stable operation of financial institutions. The problem of this research is how to find the cut-off point of the default judgment of profit maximization under the premise of ensuring the accuracy of calculating the default probability of customers. The innovation and characteristics of this article: First, the weighted average of the default probability of customers calculated by multiple different types of default prediction models ensures the overall accuracy of calculating the default probability of customers, and avoiding the inaccurate calculation of the default probability of customers using a single model. The second is to define the calculation formula of the income, loss and profit obtained by the financial institution from the loan, based on the construction principle of the ROC curve, traverse all the critical points to draw the relationship curve between the critical point and the profit, and find the judgment of the cut-off point of the default with the highest profit so as to avoid the cut-off point profit obtained by the existing methods of calculating the critical point, such as the Yordon index, generalized symmetric point estimation and empirical likelihood method. The study finds that the mixed model proposed in this paper has higher accuracy than the single model; in the Renren loan data set, the profit-driven default judgment critical point of the loan profit is much higher than that of other methods; and the virtual data, the comparative analysis of the set, shows that the default prediction model with high overall accuracy helps to mitigate the loss caused by the inappropriate selection of the default point of the default judgment in the loan process.

Key words: mixed model, default prediction, profit, cut-off point

摘要： 违约判别临界点是金融机构是否接受客户贷款申请的重要参考,合适的违约判别临界点对减少金融机构贷款损失实现稳健经营具有重要意义。本文研究的问题是如何保证计算客户违约概率的准确性,并找到利润最大化的违约判别临界点。本文的创新与特色:一是通过将多个不同类型的违约判别模型计算的客户违约概率进行加权平均,保证了计算客户违约概率的的整体准确性,避免了使用单一模型计算客户违约概率不准确的弊端;二是通过定义金融机构从贷款中获得利润的计算公式,以利润最大为目标,求解违约判别临界点,避免了现有计算临界点的方法如广义对称点估计和经验似然法等方法得到的临界点利润不是最大的弊端。研究发现:混合模型比单一模型的准确性高,AUC值显著提高;在人人贷数据集中本文的违约判别临界点下贷款利润远高于其他方法下临界点的利润。

关键词: 混合模型, 违约判别, 利润, 违约判别临界点