多决策树的模糊积分融合在银行信贷管理系统中的应用

为了提高基于数据挖掘的商业银行信贷管理系统的信贷风险评估水平，将多决策树的Choquet模糊积分融合(MTCFF)模型应用到银行信贷管理系统中。基本思想是采用决策树在已知类型的客户数据上进行挖掘，按照决策树剪枝程度不同形成不同的决策树并产生规则，利用所生成的不同决策树的规则，对未知类型的客户数据进行分类，然后让Choquet模糊积分对多棵决策树的分类结果进行融合，形成最优判断。采用UCI数据库中German客户信用卡数据集进行验证，实验证明Choquet模糊积分的非线性融合效果优于单棵决策树的分类效果，也优于其他线性融合方法，并且Choquet模糊积分要优于Sugeno模糊积分。

Application of fuzzy integral fusion of multiple decision trees into commercial bank credit management system

In order to improve the level of assessment of the credit risk of commercial bank credit management system based on data mining, the model of multiple decision trees by Choquet fuzzy integral fusion (MTCFF) was applied to the system. The basic idea was to mine the classified customer data by decision tree, form the different decision trees and rules according to different pruning degree, and detect unclassified customer data by different decision tree rules, and then nonlinearly combine the results from multiple decision trees by Choquet fuzzy integral to get the best decision. Using the German of the UCI dataset, the experimental results show that fusion of Choquet fuzzy integral is superior to the single decision tree in terms of classification accuracy, and it is also superior to other linear fusion methods. Choquet fuzzy integral is superior to Sugeno fuzzy integral.