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5 根据三个指标的集成模型准确性评估 (训练测试比为70:30)

集成算法 基分类器 准确率(%) Type-I error (%) AUC A FS A FS A FS Bagging LR 98.65 99.46 6.03 2.59 0.990 0.998 CART 97.71 98.65 6.03 5.17 0.977 0.996 C4.5 97.71 98.65 12.93 5.17 0.992 0.998 MLP 96.36 98.38 10.34 5.17 0.992 0.998 SVM 98.52 98.92 5.17 1.72 0.974 0.994 Boosting LR 97.98 98.48 5.74 3.45 0.984 0.898 CART 97.71 99.59 5.17 1.72 0.984 0.999 C4.5 97.30 98.38 6.03 3.45 0.981 0.997 MLP 97.30 99.32 7.76 3.44 0.996 0.999 SVM 98.52 98.65 4.31 2.59 0.974 0.981 Random Subspace LR 97.71 98.79 7.41 4.31 0.988 0.994 CART 96.23 97.04 12.93 5.17 0.975 0.976 C4.5 96.36 97.57 12.07 3.45 0.957 0.997 MLP 96.77 97.57 12.93 4.31 0.993 0.995 SVM 97.98 98.25 7.76 4.31 0.956 0.994 Rotation Forest LR 98.65 99.59 4.31 0.86 0.998 1.000 CART 98.11 98.65 3.44 1.72 0.993 0.995 C4.5 98.65 99.05 6.03 3.45 0.985 0.999 MLP 97.84 99.46 7.76 2.59 0.995 0.999 SVM 98.92 99.73 4.31 1.72 0.976 1.000