Table 4. SVM accuracy results (ACC ± STD) on top 20 features using 5-fold CV on different data sets.
GF-CSFS (Pineda-Bautista, Carrasco-Ochoa & Martınez-Trinidad, 2011) framework is compared against our SMBA-CSFS. FS: Fisher Score, mRMR: Minimum-Redundancy-Maximum-Relevance, MI: Mutual Information, RFS: Robust Feature Selector, EN: Elastic Net, BSL: all features. The best results are highlighted in bold. The number in parentheses is the number of features when the performance is achieved.
| Average Accuracy of top 20 features (%) | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| ALLAML | LEUKEMIA | CLL_SUB_111 | GLIOMA | LUNG_C | LUNG_D | DLBCL | CARCINOM | GCM | |
| Fisher | 95.90 ± 0.03(13) | 98.57 ± 0.03(18) | 80.41 ± 0.02(7) | 82 ± 0.16(17) | 95.09 ± 0.03(20) | 86.38 ± 0.14(16) | 100 ± 0.00(14) | 90.86 ± 0.08(20) | 98.98 ± 0.0(18) |
| Relief | 92.95 ± 0.04(5) | 95.81 ± 0.03(10) | 82.41 ± 0.05(12) | 80 ± 0.19(12) | 91.63 ± 0.02(20) | 86.39 ± 0.07(20) | 100 ± 0.00(11) | 89.68 ± 0.03(17) | 98.71 ± 0.0(20) |
| mRMR | 75.14 ± 0.09(16) | 98.57 ± 0.03(11) | 70.69 ± 0.07(12) | 62 ± 0.12(14) | 89.16 ± 0.03(20) | 86.48 ± 0.09(17) | 99.52 ± 0.01(15) | 81.61 ± 0.07(20) | 98.71 ± 0.0(20) |
| MI | 94.38 ± 0.03(18) | 97.14 ± 0.03(4) | 81.03 ± 0.05(20) | 82 ± 0.21(19) | 95.07 ± 0.015(11) | 79.90 ± 0.18(14) | 100 ± 0.00(19) | 90.86 ± 0.06(11) | 98.67 ± 0.0(19) |
| ls-21 | 76.47 ± 0.13(6) | 65.52 ± 0.08(3) | 63.44 ± 0.03(20) | 46 ± 0.21(7) | 73.88 ± 0.04(19) | 75.43 ± 0.07(18) | 93.46 ± 0.03(20) | 39.68 ± 0.04(19) | 97.59 ± 0.0(19) |
| ll-21 | 82.1 ± 0.05(16) | 80.67 ± 0.09(15) | 74.58 ± 0.07(20) | 68 ± 0.13(18) | 91.15 ± 0.02(15) | 67.24 ± 0.12(15) | 96.38 ± 0.02(17) | 72.40 ± 0.05(17) | 96.87 ± 0.0(20) |
| RFS | 79.24 ± 0.168(17) | 74.95 ± 0.09(6) | 71.94 ± 0.10(19) | 68 ± 0.21(13) | 82.79 ± 0.05(17) | 68.67 ± 0.07(18) | 96.62 ± 0.01(20) | 58.03 ± 0.18(20) | 96.97 ± 0.01(20) |
| LASSO | 95.73 ± 0.02(6) | 70.3 ± 0.08(15) | 71.29 ± 0.05(18) | 81.67 ± 0.08(19) | 96.26 ± 0.00(18) | 93.22 ± 0.021(20) | 100 ± 0.00(10) | 87.88 ± 0.03(18) | 96.09 ± 0.0(20) |
| EN | 95.73 ± 0.04(10) | 70.3 ± 0.08(15) | 68.73 ± 0.10(19) | 81.67 ± 0.08(19) | 95.97 ± 0.012(18) | 93.22 ± 0.021(20) | 100 ± 0.00(10) | 88.56 ± 0.03(19) | 96.09 ± 0.0(20) |
| SMBA-CSFS | 88.24 ± 0.04(20) | 81.93 ± 0.02(20) | 75.53 ± 0.06(20) | 73.34 ± 0.18(16) | 98.41 ± 0.014(19)} | 97.93 ± 0.03(19) | 98.30 ± 0.02(13) | 94.95 ± 0.02(19) | 99.2 ± 0.01(20) |
| BSL | 97.89 ± 0.04 | 98.95 ± 0.021 | 84.26 ± 0.06 | 85 ± 0.1 | 99.57 ± 0.00 | 98.62 ± 0.02 | 100 ± 0.00 | 98.65 ± 0.01 | 100 ± 0.00 |