Table 17 Results of some simple majority vote ensembles and comparison with best single methods in terms of F-score
From: A detailed error analysis of 13 kernel methods for protein-protein interaction extraction
Combination | Corpus | P | R | F |
|---|---|---|---|---|
Single best | ||||
APG | AIMed | 59.9 | 53.6 | 56.2 |
APG | BioInfer | 60.2 | 61.3 | 60.7 |
kBSPS | HPRD50 | 60.0 | 88.4 | 70.2 |
APG | IEPA | 66.6 | 82.6 | 73.1 |
kBSPS | LLL | 69.9 | 95.9 | 79.3 |
APG+SL+kBSPS | AIMed | 58.0 | 61.1 | 58.9 |
BioInfer | 60.3 | 66.4 | 63.0 | |
HPRD50 | 67.6 | 76.9 | 71.4 | |
IEPA | 68.6 | 85.3 | 75.4 | |
LLL | 71.7 | 94.5 | 80.0 | |
APG+SL+BayesNet | AIMed | 55.9 | 60.3 | 57.6 |
BioInfer | 58.6 | 68.8 | 63.1 | |
HPRD50 | 68.4 | 69.8 | 67.7 | |
IEPA | 71.1 | 79.9 | 74.5 | |
LLL | 74.3 | 92.9 | 80.8 | |
All 13 kernels | AIMed | 67.5 | 35.8 | 46.6 |
BioInfer | 61.5 | 56.5 | 58.7 | |
HPRD50 | 65.4 | 69.3 | 66.1 | |
IEPA | 70.5 | 78.8 | 73.7 | |
LLL | 69.6 | 98.7 | 79.5 |