Table 5.
Experimental results on Statlog dataset with the best feature subset
| Mean ± SD | RF | LR | SVM | ELM | KNN | Proposed ensemble |
|---|---|---|---|---|---|---|
| E (%) | 87.53 ± 5.39 | 87.87 ± 6.82 | 88.67 ± 5.02 | 82.81 ± 5.54 | 76.94 ± 11.33 | 94.44 ± 3.78 |
| Precision (%) | 83.70 ± 6.58 | 84.07 ± 8.01 | 84.81 ± 6.40 | 78.15 ± 6.64 | 70 ± 15.37 | 92.59 ± 4.62 |
| Recall (%) | 80.64 ± 11.80 | 82.08 ± 13.07 | 83.85 ± 10.98 | 70.65 ± 13.77 | 62.85 ± 17.51 | 92.15 ± 7.10 |
| G-mean | 83.14 ± 7.54 | 83.79 ± 8.19 | 84.41 ± 7.10 | 76.65 ± 8.05 | 68.40 ± 15.63 | 92.56 ± 4.79 |
| MC (%) | 51.85 ± 26.07 | 50 ± 34.67 | 44.81 ± 23.78 | 75.19 ± 29.63 | 96.67 ± 44.56 | 22.22 ± 19.36 |
| Specificity (%) | 86.13 ± 6.17 | 86 ± 6.58 | 85.45 ± 7.83 | 84.29 ± 8.43 | 75.18 ± 16.55 | 93.21 ± 5.43 |
| AUC (%) | 83.75 ± 8.26 | 83.92 ± 9.44 | 85.07 ± 7.72 | 80.17 ± 6.96 | 68.42 ± 13.73 | 92.08 ± 5.51 |
The average sd on 10-folds CV. The best result is bolded