Table 2 Best testing accuracy and standard errors (mean ± standard error, %) with classification models derived from best training, with the use of GLGS and SVMRFE feature selection algorithms and seven learning classifiers. By using each feature selection algorithm on each data set, the best result as well as the classifier is highlighted in bold.
From: Comparison of feature selection and classification for MALDI-MS data
Learning classifier | GLGS | SVMRFE | ||||
|---|---|---|---|---|---|---|
| Â | Ovarian cancer | Breast cancer | Liver disease | Ovarian cancer | Breast cancer | Liver disease |
KNNC | 88.0 ± 5.8% | 80.5 ± 8.6 | 88.3 ± 6.3 | 96.6 ± 2.9 | 87.9 ± 7.0 | 95.3 ± 3.4 |
NBC | 79.9 ± 5.3 | 75.8 ± 9.0 | 90.8 ± 5.6 | 90.9 ± 4.5 | 76.0 ± 9.1 | 96.5 ± 3.7 |
NMSC | 82.6 ± 5.1 | 77.8 ± 9.1 | 92.1 ± 4.4 | 92.6 ± 3.8 | 81.8 ± 7.6 | 96.5 ± 4.0 |
UDC | 82.7 ± 5.4 | 78.0 ± 8.0 | 91.3 ± 5.6 | 92.5 ± 4.4 | 82.4 ± 7.7 | 91.7 ± 5.8 |
SVM_linear | 89.6 ± 4.9 | 85.6 ± 8.3 | 95.8 ± 3.8 | 97.9 ± 2.0 | 89.9 ± 6.0 | 98.2 ± 2.7 |
SVM_rbf | 90.4 ± 4.3 | 85.3 ± 7.9 | 96.4 ± 3.3 | 98.2 ± 1.8 | 90.5 ± 6.1 | 97.5 ± 3.1 |
LMNN | 93.1 ± 4.4 | 88.3 ± 7.4 | 97.4 ± 3.2 | 99.2 ± 1.1 | 91.7 ± 4.5 | 99.0 ± 1.8 |