基于异类类内超平面的模糊支持向量机及其应用

针对如何基于不平衡信贷数据对借贷人的信用进行合理准确地评估问题，基于个人信用的统计数据，提出了一种新的个人贷款信用风险的评估方法。首先，构建了个人贷款信用风险评估指标体系，结合IV模型，对各特征进行重要性分析。其次，结合模糊数学理论，设计了一种基于异类类内超平面的隶属函数。最后，结合支持向量机，构建了一种新的个人贷款信用风险评估方法——基于异类类内超平面的模糊支持向量机。结果表明：个人贷款信用风险评估的可用额度比值、逾期30~59天的次数、逾期90天及以上次数以及逾期60~89天次数这4个指标的IV值均大于0.3，重要性较强，表明对贷款人的信用评估影响较大；所设计的隶属函数能对不同样本赋予不同权重，可充分体现不同样本的重要性；基于异类类内超平面的模糊支持向量机在一定程度上可以提高贷款人的信用风险评估精度，证明了该方法的有效性和可行性。

Fuzzy Support Vector Machine Based on Heterogeneous Class Internal Hyperplane and Its Application

Aiming at the problem of how to reasonably and accurately evaluate the credit of the borrower based on the unbalanced credit data, a new method for evaluating the personal credit risk was proposed. Firstly, the personal credit risk assessment index system was constructed, and in order to analyze the importance of each characteristic of the data, IV values of the credit characteristics of the data were calculated with IV model. Secondly, combining with fuzzy mathematics theory, a membership function based on the heterogeneous class hyperplane was designed. At last, combining the traditional support vector machine (SVM), a fuzzy support vector machine (FSVM) based on the heterogeneous class hyperplane was constructed. The results show that IV values of four indexes of personal loan credit risk assessment:, namely the available amount ratio, the number of overdue 30 to 59 days, the number of overdue 90 days or more, and the number of overdue 60 to 89 days are all more than 0.3, which are of great importance, indicating that they have a great impact on the credit assessment of the lender. The constructed membership function can provide different weights for different samples, and can reflect the importance of different samples. The constructed FSVM based on the heterogeneous class hyperplane can effectively improve the accuracy of estimating credit risk for personal loans, which shows the feasibility and the effectiveness of the proposed method.