Table 5 Results of different classifiers without/with PCA in Kaggle 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 | |
|---|---|---|---|---|---|---|
No Feature extraction (PCA feature extraction) | Single linear classifier | LogR | 0.6329 (0.6219) | 0.6370 (0.6229) | 0.6321 (0.6232) | 0.6792 (0.6761) |
LDA | 0.6081 (0.6200) | 0.6500 (0.6194) | 0.5673 (0.6233) | 0.6466 (0.6791) | ||
Single nonlinear classifier | KNN | 0.5714 (0.5500) | 0.7080 (0.6887) | 0.4390 (0.4108) | 0.6156 (0.5901) | |
SVM | 0.6381 (0.6271) | 0.6380 (0.6279) | 0.6420 (0.6311) | 0.6935 (0.6873) | ||
BPNN | 0.6081 (0.5586) | 0.5584 (0.5150) | 0.6604 (0.6018) | 0.6537 (0.5859) | ||
CART | 0.5786 (0.5376) | 0.5784 (0.5401) | 0.5808 (0.5361) | 0.5896 (0.5436) | ||
Linear ensemble classifier | LogR Bagging | 0.6257 (0.6224) | 0.6376 (0.6401) | 0.6162 (0.6064) | 0.6727 (0.6687) | |
LDA Bagging | 0.6171 (0.6195) | 0.6348 (0.6516) | 0.5988 (0.5893) | 0.6457 (0.6731) | ||
Nonlinear ensemble classifier | KNN Bagging | 0.5700 (0.5586) | 0.6852 (0.6969) | 0.4584 (0.4197) | 0.6170 (0.5909) | |
SVM Bagging | 0.5819 (0.6281) | 0.5314 (0.6616) | 0.6433 (0.5960) | 0.6401 (0.6771) | ||
BPNN Bagging | 0.6343 (0.5986) | 0.6228 (0.6140) | 0.6505 (0.5870) | 0.6709 (0.6497) | ||
CART Bagging | 0.5890 (0.5495) | 0.5951 (0.5552) | 0.5847 (0.5446) | 0.5993 (0.5734) | ||