Table 2. Comparison of metabolite selection strategies.
3-week data | 6-week data | Liver profile | |
EFM | 1 | .78 | 0.81 |
1-Neighborhood | 0.93 | 0.63 | 0.72 |
Random | 0.68 | 0.68 | 0.81 |
Classification accuracies of each metabolite selection strategy per dataset are shown. For the random selection case, the number of subsets to consider is matched to the highest number of datasets among the competitors. Results show that the EFM strategy weakly dominates the competitors.