From: A kernel-based approach for detecting outliers of high-dimensional biological data
Data set | Measurements | Without outlier removal | After outlier removal | ||
---|---|---|---|---|---|
 |  |  | KLOD | Mahalanobis | One-class SVM |
Leukemia | Specificity (%) | 96.17 | 99.00 | 97.37 | 100 |
 | Sensitivity (%) | 95.60 | 99.44 | 100 | 95.24 |
 | Accuracy (%) | 95.97 | 99.13 | 98.28 | 98.33 |
 | No. of the outliers | ALL | 2 | 9 | 8 |
 |  | AML | 7 | 5 | 4 |
Colon | Specificity (%) | 82.50 | 85.95 | 83.25 | 85.26 |
 | Sensitivity (%) | 88.25 | 94.43 | 85.90 | 94.17 |
 | Accuracy (%) | 86.21 | 91.25 | 85.00 | 91.09 |
 | No. of the outliers | normal | 1 | 2 | 3 |
 |  | tumor | 5 | 1 | 4 |