Table 2. Overall comparison to single-expert classifiers.
| Disease group | Techniques | Precision (p) | Recall (r) | F-measure (F) |
| Cardiovascular | MSVM | 74.3% | 87.6% | 80.4% |
| WNB | 57.3% | 72.5% | 63.9% | |
| WKNN(3) | 60.1% | 68.6% | 64.0% | |
| EPU | 85.2% | 81.0% | 84.1% | |
| Endocrine | MSVM | 83.4% | 85.2% | 84.2% |
| WNB | 61.3% | 70.4% | 65.3% | |
| WKNN(3) | 64.5% | 53.1% | 57.9% | |
| EPU | 88.1% | 87.7% | 87.9% | |
| Neurological | MSVM | 69.3% | 83.7% | 75.8% |
| WNB | 61.1% | 74.4% | 67.0% | |
| WKNN(3) | 62.3% | 67.1% | 64.6% | |
| EPU | 78.2% | 80.4% | 78.6% | |
| Metabolic | MSVM | 84.0% | 91.3% | 87.4% |
| WNB | 68.8% | 79.9% | 73.9% | |
| WKNN(3) | 76.6% | 78.8% | 77.6% | |
| EPU | 83.3% | 93.9% | 90.9% | |
| Ophthalmological | MSVM | 78.4% | 86.1% | 81.9% |
| WNB | 61.2% | 78.7% | 68.8% | |
| WKNN(3) | 67.3% | 72.2% | 69.6% | |
| EPU | 89.3% | 81.0% | 84.7% | |
| Cancer | MSVM | 73.4% | 83.9% | 78.3% |
| WNB | 72.5% | 85.1% | 78.3% | |
| WKNN(3) | 76.4% | 81.0% | 78.6% | |
| EPU | 81.2% | 84.5% | 82.6% | |
| Average performance | MSVM | 78.6% | 86.3% | 81.3% |
| WNB | 63.7% | 76.8% | 69.5% | |
| WKNN(3) | 67.9% | 70.1% | 68.7% | |
| EPU | 84.2% | 84.8% | 84.8% |
EPU is compared with its three component classifiers Multi-level Support Vector Machine (MSVM), Weighted Naïve Bayes (WNB) and Weighted K-Nearest Neighbor (KNN) on 6 disease groups. WKNN(3) is an instance-based classifier that predicts the class of an unlabeled gene based on its 3 closest labeled genes.