Table 2 Results of different classifiers with/without PCA in China Unionpay dataset when 100 < feature dimensions < sample size
From: A high-dimensionality-trait-driven learning paradigm for high dimensional credit classification
Feature extraction | Classifer | Total | TP | TN | AUC | |
|---|---|---|---|---|---|---|
PCA-based feature extraction (Non-feature extraction) | Single linear classifier | LogR | 0.6965 (0.6959) | 0.7473 (0.7348) | 0.6459 (0.6571) | 0.7481 (0.7463) |
LDA | 0.6982 (0.6696) | 0.7449 (0.7447) | 0.6517 (0.5944) | 0.7477 (0.7137) | ||
Single nonlinear classifier | KNN | 0.6468 (0.6699) | 0.7475 (0.7237) | 0.5462 (0.6163) | 0.7020 (0.7208) | |
SVM | 0.7024 (0.7029) | 0.7479 (0.6926) | 0.6571 (0.7134) | 0.7574 (0.7561) | ||
BPNN | 0.6890 (0.6860) | 0.7111 (0.7091) | 0.6672 (0.6626) | 0.7355 (0.7321) | ||
CART | 0.6103 (0.6277) | 0.6084 (0.6304) | 0.6123 (0.6250) | 0.6039 (0.6349) | ||
Linear ensemble classifier | LogR Bagging | 0.6970 (0.6912) | 0.7500 (0.7345) | 0.6440 (0.6480) | 0.7472 (0.7430) | |
LDA Bagging | 0.6954 (0.6697) | 0.7415 (0.7475) | 0.6496 (0.5918) | 0.7455 (0.7154) | ||
Nonlinear ensemble classifier | KNN Bagging | 0.6100 (0.6581) | 0.7612 (0.7258) | 0.4594 (0.5903) | 0.6638 (0.7103) | |
SVM Bagging | 0.6618 (0.6489) | 0.6836 (0.6219) | 0.6404 (0.6757) | 0.7005 (0.7062) | ||
BPNN Bagging | 0.7005 (0.6900) | 0.7075 (0.7062) | 0.6940 (0.6739) | 0.7434 (0.7364) | ||
CART Bagging | 0.6484 (0.5446) | 0.6357 (0.5540) | 0.6614 (0.5370) | 0.6600 (0.5583) | ||