Table 9.
Accuracy of the k-NN classifier trained on the ENDM-based filtered training set with the noise correction modality, four different metrics (SuMax, UnMax, SuSum, UnSum) and three different classifiers (Bagging, AdaBoost, and k-NN with ). The values in brackets are the detected noise ratio. The best results are marked in bold.
| SuMax | UnMax | SuSum | UnSum | ||
|---|---|---|---|---|---|
| Letter | Bagging | 74.10 (1%) | 73.96 (1%) | 73.02 (0%) | 74.94 (2%) |
| AdaBoost | 73.74 (1%) | 73.74 (1%) | 72.86 (0%) | 73.76 (1%) | |
| k-NN | 76.44 (11%) | 75.72 (4%) | 80.44 (15%) | 75.98 (5%) | |
| Optdigits | Bagging | 76.60 (0%) | 76.60 (0%) | 78.80 (2%) | 76.60 (0%) |
| AdaBoost | 88.90 (15%) | 89.30 (17%) | 90.10 (19%) | 88.00 (17%) | |
| k-NN | 89.10 (16%) | 89.20 (17%) | 90.40 (20%) | 88.00 (20%) | |
| Pendigit | Bagging | 83.55 (3%) | 84.25 (4%) | 79.90 (0%) | 83.55 (3%) |
| AdaBoost | 93.15 (21%) | 88.80 (9%) | 94.15 (19%) | 91.85 (13%) | |
| k-NN | 92.60 (15%) | 91.75 (15%) | 93.65 (18%) | 92.05 (16%) | |
| Statlog | Bagging | 80.30 (8%) | 84.55 (16%) | 73.70 (1%) | 82.20 (10%) |
| AdaBoost | 83.40 (11%) | 82.40 (10%) | 85.75 (18%) | 83.80 (12%) | |
| k-NN | 85.05 (18%) | 84.45 (16%) | 85.80 (19%) | 84.70 (16%) | |
| Vehicle | Bagging | 63.00 (2%) | 63.00 (2%) | 60.00 (1%) | 62.00 (1%) |
| AdaBoost | 65.50 (15%) | 65.50 (10%) | 67.50 (20%) | 66.00 (15%) | |
| k-NN | 64.00 (12%) | 63.50 (8%) | 64.50 (6%) | 61.50 (2%) |