Table 8 Comparisons of NRF values between PhyloMissForest and the baseline algorithm MissForest [37]. “Split-LS-Rand” refers to the configuration where LS is incorporated for guidance purposes in the different steps of PhyloMissForest, while “Best Observed” represents the best results reported with any possible configuration of search strategies. Lower values denote better quality
From: PhyloMissForest: a random forest framework to construct phylogenetic trees with missing data
Dataset | %Missing | PhyloMissForest | Mixed-type | |
|---|---|---|---|---|
Split-LS-Rand | Best Observed | MissForest | ||
9x9 | 5% | 4.6 | 2.9 | 5.0 |
10% | 7.5 | 6.7 | 11.7 | |
15% | 7.5 | 7.5 | 12.1 | |
20% | 12.5 | 9.2 | 13.8 | |
25% | 12.5 | 12.5 | 19.2 | |
30% | 12.9 | 12.9 | 17.9 | |
37x37 | 5% | 1.2 | 1.2 | 2.0 |
10% | 2.8 | 2.8 | 4.6 | |
15% | 4.6 | 4.6 | 7.0 | |
20% | 6.8 | 6.5 | 9.4 | |
25% | 9.3 | 9.3 | 14.0 | |
30% | 13.2 | 13.2 | 16.2 | |
55x55 | 5% | 3.4 | 3.0 | 6.3 |
10% | 5.9 | 5.7 | 8.7 | |
15% | 11.0 | 10.1 | 14.7 | |
20% | 14.3 | 14.3 | 16.9 | |
25% | 20.5 | 19.7 | 22.0 | |
30% | 22.3 | 22.3 | 25.0 | |