基于集成学习的中国P2P网络借贷信用风险预警模型的对比研究*

操玮, 李灿, 贺婷婷, 朱卫东

Predicting Credit Risks of P2P Loans in China Based on Ensemble Learning Methods

Cao Wei,Li Can,He Tingting,Zhu Weidong

表6 根据三个指标的集成模型准确性评估 (训练测试比为80:20)

集成算法 基分类器 准确率(%) Type-I error (%) AUC A FS A FS A FS Bagging LR 97.17 98.99 7.32 4.90 0.962 0.990 CART 97.17 98.78 8.53 6.09 0.968 0.994 C4.5 97.57 98.58 9.76 4.90 0.995 0.997 MLP 96.56 98.79 8.53 6.09 0.989 0.985 SVM 96.37 98.38 9.76 3.66 0.976 0.986 Boosting LR 96.56 97.57 13.41 6.09 0.966 0.995 CART 97.37 97.77 8.53 3.66 0.995 0.997 C4.5 95.14 97.36 8.53 4.90 0.972 0.980 MLP 97.37 98.58 9.76 6.09 0.993 0.996 SVM 97.37 98.18 7.32 3.66 0.982 0.990 Random Subspace LR 94.13 97.36 20.73 8.53 0.966 0.994 CART 96.56 96.96 14.63 9.76 0.940 0.997 C4.5 96.56 97.16 12.20 10.98 0.964 0.975 MLP 95.95 97.97 8.53 7.32 0.968 0.991 SVM 96.56 97.77 14.63 9.76 0.967 0.983 Rotation Forest LR 98.38 99.39 6.09 1.20 0.994 1.000 CART 97.57 99.19 6.09 1.20 0.992 0.996 C4.5 97.57 99.19 7.32 1.20 0.981 0.998 MLP 98.38 99.19 7.32 2.44 0.992 1.000 SVM 98.58 99.39 6.09 1.20 0.981 1.000