Table 6 Results of different classifiers without/with PCA in China Unionpay dataset when feature dimensions ≥ sample size
From: A high-dimensionality-trait-driven learning paradigm for high dimensional credit classification
Feature extraction | Classifier | Total | TP | TN | AUC | |
|---|---|---|---|---|---|---|
LDA | 0.5220 (0.6220) | 0.4730 (0.6597) | 0.5907 (0.5836) | 0.6577 (0.6534) | ||
Single nonlinear classifier | KNN | 0.6559 (0.5932) | 0.7535 (0.8672) | 0.5521 (0.3004) | 0.7145 (0.6635) | |
SVM | 0.6627 (0.6085) | 0.7111 (0.7441) | 0.6183 (0.4711) | 0.7014 (0.6726) | ||
BPNN | 0.6017 (0.5831) | 0.7474 (0.7183) | 0.4408 (0.4239) | 0.6096 (0.6022) | ||
CART | 0.5780 (0.5644) | 0.5601 (0.5962) | 0.5957 (0.5314) | 0.5958 (0.5686) | ||
Linear ensemble classifier | LogR Bagging | 0.6407 (0.6237) | 0.6562 (0.6787) | 0.6287 (0.5704) | 0.6532 (0.6576) | |
LDA Bagging | 0.6729 (0.6271) | 0.7594 (0.6847) | 0.5828 (0.5738) | 0.6981 (0.6733) | ||
Nonlinear ensemble classifier | KNN Bagging | 0.6356 (0.5627) | 0.7996 (0.8975) | 0.4634 (0.2069) | 0.6868 (0.6630) | |
SVM Bagging | 0.6695 (0.6305) | 0.6824 (0.6977) | 0.6570 (0.5549) | 0.7158 (0.6798) | ||
BPNN Bagging | 0.6576 (0.5949) | 0.7621 (0.6747) | 0.5379 (0.5050) | 0.6892 (0.6522) | ||
CART Bagging | 0.5542 (0.5831) | 0.4748 (0.6842) | 0.6371 (0.4779) | 0.5665 (0.5547) | ||