基于ADASYN-AdaBoost-CNN的信用风险评估模型

Credit Risk Assessment Model Based on Adasyn-Adaboost-CNN

数据失衡对信用风险评估模型的性能构成挑战,为了提高金融机构的风险控制能力,针对信用风险数据的不平衡现象,提出了一种基于ADASYN-AdaBoost-CNN的集成模型。首先采用ADASYN技术平衡数据集,减轻不平衡现象对信用风险评估的影响。其次构建基于卷积神经网络的集成学习算法,确保信用风险评估模型的准确性与鲁棒性。最后在lendingclub借贷数据集上开展实验,使用F1-measure、G-mean和AUC三个评价指标检验模型性能。结果表明,ADASYN-AdaBoost-CNN模型能够有效解决不平衡信用风险评估问题。

Data imbalance poses a challenge to the performance of credit risk assessment model.In order to improve the risk con⁃trol ability of financial institutions,an integrated model based on ADASYN-AdaBoost-CNN is proposed solve the imbalance phenome⁃non of credit risk data.Firstly,ADASYN technique is used to balance the data set to reduce the impact of the imbalance on credit risk assessment.Secondly,an ensemble learning algorithm based on convolutional neural network is constructed to ensure the accuracy and robustness of the credit risk assessment model.Finally,experiments are carried out on Lending Club lending data set,and three evalua⁃tion indexes,F1-measure,G-mean and AUC,are used to test the performance of the model.The results show that the ADASYN-Ada⁃Boost-CNN model can effectively solve the problem of unbalanced credit risk assessment.