Performance evaluation with AUC (A), precision (P), and recall (R) (with the standard mean error in brackets)
| Comparison . | Method . | EW-T2D . | LC . | C-T2D . | IBD . | Obesity . | |
|---|---|---|---|---|---|---|---|
| A. Taxonomic features | RF | A | 78.2%(3.2%) | 95.1%(1.5%) | 75.1%(2.6%) | 92.6%(2.2%) | 65.3%(4.4%) |
| P | 71.3%(2.3%) | 92.6%(1.7%) | 66.9%(1.9%) | 70.0%(20.0%) | 64.1%(0.4%) | ||
| R | 80.0%(3.4%) | 82.5%(3.6%) | 62.9%(3.9%) | 20.0%(6.3%) | 97.6%(1.5%) | ||
| SVM | A | 75.5%(3.1%) | 89.4%(2.2%) | 71.6%(2.4%) | 92.0%(2.6%) | 64.2%(3.8%) | |
| P | 75.2%(3.6%) | 81.9%(3.3%) | 65.7%(3.5%) | 66.7%(18.3%) | 70.9%(2.5%) | ||
| R | 69.1%(6.2%) | 75.0%(4.2%) | 61.8%(4.7%) | 36.0%(9.8%) | 77.0%(3.1%) | ||
| MLP | A | 74.1%(5.1%) | 90.0%(1.5%) | 72.3%(2.2%) | 86.4%(5.0%) | 52.3%(3.0%) | |
| P | 86.4%(6.9%) | 79.3%(2.1%) | 67.3%(2.0%) | 86.3%(2.3%) | 67.2%(1.5%) | ||
| R | 69.1%(6.8%) | 86.1%(3.2%) | 64.1%(5.8%) | 94.1%(2.6%) | 70.3%(1.5%) | ||
| FT-Transformer | A | 78.4%(3.6%) | 90.4%(2.4%) | 73.2%(3.0%) | 93.6%(2.0%) | 66.1%(1.7%) | |
| P | 71.4%(4.7%) | 84.8%(6.2%) | 66.8%(3.2%) | 70.1%(7.7%) | 75.9%(2.5%) | ||
| R | 69.1%(6.8%) | 85.8%(3.4%) | 64.1%(3.3%) | 72.0%(4.9%) | 67.3%(6.9%) | ||
| DeepMicro [7] | A | 82.9%(3.9%) | 88.8%(1.1%) | 72.5%(2.5%) | 87.3%(3.0%) | 67.4%(3.4%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| EnsDeepDP [4] | A | 86.7%(N/A) | 94.3%(N/A) | 77.6%(N/A) | 95.8%(N/A) | 72.3%(N/A) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| B. Taxonomic features with external knowledge | Meta-Signer [15] | A | – | 90.5%(5.0%) | – | 79.4(15.9%) | 60.0%(13.5%) |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| TaxoNN [16] | A | – | 91.1%(N/A) | 73.3%(N/A) | – | – | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| PopPhy-CNN [14] | A | – | 90.1%(N/A) | – | – | 58.9%(N/A) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| MicroKPNN [18] | A | 85.8(6.7%) | 96.9%(0.9%) | 75.5%(3.2%) | 95.4%(3.7%) | 72.8%(4.8%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| C. Taxonomic features and functional features | MDL4Microbiome | A | 76.8%(5.7%) | 92.5%(1.3%) | 74.7%(1.3%) | 88.0%(0.6%) | 54.1%(2.0%) |
| P | 76.3%(7.5%) | 78.1%(1.4%) | 67.6%(1.9%) | 86.0%(1.7%) | 69.3%(1.9%) | ||
| R | 69.1%(4.6%) | 89.1%(5.4%) | 66.5%(4.2%) | 92.9%(2.9%) | 72.1%(3.2%) | ||
| MVIB [29] | A | 85.9%(2.3%) | 92.5%(0.5%) | 75.8%(1.2%) | 93.6%(1.4%) | 66.6%(2.7%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| FT-Concat | A | 82.6%(3.1%) | 91.7%(1.7%) | 77.5%(1.9%) | 92.0%(2.4%) | 61.5%(4.2%) | |
| P | 77.8%(4.0%) | 89.4%(3.2%) | 68.9%(2.2%) | 79.3%(9.7%) | 75.1%(3.3%) | ||
| R | 80.0%(7.3%) | 76.7%(4.5%) | 68.2%(6.1%) | 64.0%(7.5%) | 57.6%(12.4%) | ||
| FT-Vote | A | 82.2%(2.1%) | 91.4%(1.4%) | 76.6%(1.6%) | 94.6%(1.7%) | 61.0%(3.3%) | |
| P | 70.5%(5.4%) | 84.4%(4.0%) | 68.3%(1.5%) | 77.7%(6.1%) | 70.4%(2.1%) | ||
| R | 80.0%(6.0%) | 87.5%(2.9%) | 63.5%(4.3%) | 56.0%(11.7%) | 63.0%(11.9%) | ||
| T-MBT | A | 83.2%(4.2%) | 91.7%(2.3%) | 76.6%(1.6%) | 95.5%(2.1%) | 62.2%(3.3%) | |
| P | 80.0%(2.8%) | 87.1%(4.2%) | 69.7%(2.8%) | 81.7%(7.6%) | 73.1%(2.6%) | ||
| R | 67.3%(6.2%) | 80.8%(6.0%) | 70.0%(1.7%) | 68.0%(8.0%) | 60.6%(7.3%) | ||
| MSFT-Transformer | A | 87.5%(4.1%) | 94.1%(1.2%) | 77.6%(1.6%) | 97.4%(1.8%) | 67.7%(2.9%) | |
| P | 80.5%(7.0%) | 88.7%(3.9%) | 70.8%(2.8%) | 100.0%(0.0%) | 72.7%(2.9%) | ||
| R | 83.6%(6.0%) | 89.2%(3.4%) | 65.3%(2.5%) | 72.0%(13.6%) | 69.7%(10.5%) | ||
| Comparison . | Method . | EW-T2D . | LC . | C-T2D . | IBD . | Obesity . | |
|---|---|---|---|---|---|---|---|
| A. Taxonomic features | RF | A | 78.2%(3.2%) | 95.1%(1.5%) | 75.1%(2.6%) | 92.6%(2.2%) | 65.3%(4.4%) |
| P | 71.3%(2.3%) | 92.6%(1.7%) | 66.9%(1.9%) | 70.0%(20.0%) | 64.1%(0.4%) | ||
| R | 80.0%(3.4%) | 82.5%(3.6%) | 62.9%(3.9%) | 20.0%(6.3%) | 97.6%(1.5%) | ||
| SVM | A | 75.5%(3.1%) | 89.4%(2.2%) | 71.6%(2.4%) | 92.0%(2.6%) | 64.2%(3.8%) | |
| P | 75.2%(3.6%) | 81.9%(3.3%) | 65.7%(3.5%) | 66.7%(18.3%) | 70.9%(2.5%) | ||
| R | 69.1%(6.2%) | 75.0%(4.2%) | 61.8%(4.7%) | 36.0%(9.8%) | 77.0%(3.1%) | ||
| MLP | A | 74.1%(5.1%) | 90.0%(1.5%) | 72.3%(2.2%) | 86.4%(5.0%) | 52.3%(3.0%) | |
| P | 86.4%(6.9%) | 79.3%(2.1%) | 67.3%(2.0%) | 86.3%(2.3%) | 67.2%(1.5%) | ||
| R | 69.1%(6.8%) | 86.1%(3.2%) | 64.1%(5.8%) | 94.1%(2.6%) | 70.3%(1.5%) | ||
| FT-Transformer | A | 78.4%(3.6%) | 90.4%(2.4%) | 73.2%(3.0%) | 93.6%(2.0%) | 66.1%(1.7%) | |
| P | 71.4%(4.7%) | 84.8%(6.2%) | 66.8%(3.2%) | 70.1%(7.7%) | 75.9%(2.5%) | ||
| R | 69.1%(6.8%) | 85.8%(3.4%) | 64.1%(3.3%) | 72.0%(4.9%) | 67.3%(6.9%) | ||
| DeepMicro [7] | A | 82.9%(3.9%) | 88.8%(1.1%) | 72.5%(2.5%) | 87.3%(3.0%) | 67.4%(3.4%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| EnsDeepDP [4] | A | 86.7%(N/A) | 94.3%(N/A) | 77.6%(N/A) | 95.8%(N/A) | 72.3%(N/A) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| B. Taxonomic features with external knowledge | Meta-Signer [15] | A | – | 90.5%(5.0%) | – | 79.4(15.9%) | 60.0%(13.5%) |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| TaxoNN [16] | A | – | 91.1%(N/A) | 73.3%(N/A) | – | – | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| PopPhy-CNN [14] | A | – | 90.1%(N/A) | – | – | 58.9%(N/A) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| MicroKPNN [18] | A | 85.8(6.7%) | 96.9%(0.9%) | 75.5%(3.2%) | 95.4%(3.7%) | 72.8%(4.8%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| C. Taxonomic features and functional features | MDL4Microbiome | A | 76.8%(5.7%) | 92.5%(1.3%) | 74.7%(1.3%) | 88.0%(0.6%) | 54.1%(2.0%) |
| P | 76.3%(7.5%) | 78.1%(1.4%) | 67.6%(1.9%) | 86.0%(1.7%) | 69.3%(1.9%) | ||
| R | 69.1%(4.6%) | 89.1%(5.4%) | 66.5%(4.2%) | 92.9%(2.9%) | 72.1%(3.2%) | ||
| MVIB [29] | A | 85.9%(2.3%) | 92.5%(0.5%) | 75.8%(1.2%) | 93.6%(1.4%) | 66.6%(2.7%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| FT-Concat | A | 82.6%(3.1%) | 91.7%(1.7%) | 77.5%(1.9%) | 92.0%(2.4%) | 61.5%(4.2%) | |
| P | 77.8%(4.0%) | 89.4%(3.2%) | 68.9%(2.2%) | 79.3%(9.7%) | 75.1%(3.3%) | ||
| R | 80.0%(7.3%) | 76.7%(4.5%) | 68.2%(6.1%) | 64.0%(7.5%) | 57.6%(12.4%) | ||
| FT-Vote | A | 82.2%(2.1%) | 91.4%(1.4%) | 76.6%(1.6%) | 94.6%(1.7%) | 61.0%(3.3%) | |
| P | 70.5%(5.4%) | 84.4%(4.0%) | 68.3%(1.5%) | 77.7%(6.1%) | 70.4%(2.1%) | ||
| R | 80.0%(6.0%) | 87.5%(2.9%) | 63.5%(4.3%) | 56.0%(11.7%) | 63.0%(11.9%) | ||
| T-MBT | A | 83.2%(4.2%) | 91.7%(2.3%) | 76.6%(1.6%) | 95.5%(2.1%) | 62.2%(3.3%) | |
| P | 80.0%(2.8%) | 87.1%(4.2%) | 69.7%(2.8%) | 81.7%(7.6%) | 73.1%(2.6%) | ||
| R | 67.3%(6.2%) | 80.8%(6.0%) | 70.0%(1.7%) | 68.0%(8.0%) | 60.6%(7.3%) | ||
| MSFT-Transformer | A | 87.5%(4.1%) | 94.1%(1.2%) | 77.6%(1.6%) | 97.4%(1.8%) | 67.7%(2.9%) | |
| P | 80.5%(7.0%) | 88.7%(3.9%) | 70.8%(2.8%) | 100.0%(0.0%) | 72.7%(2.9%) | ||
| R | 83.6%(6.0%) | 89.2%(3.4%) | 65.3%(2.5%) | 72.0%(13.6%) | 69.7%(10.5%) | ||
The reported results of cited methods are based on previous works that use the same datasets. There are a total of nine methods (uncited) evaluated with our framework for five trails, including MSFT-Transformer. N/A indicates that the corresponding study did not provide the standard mean error. The best performance of each dataset is highlighted in bold while the second-high underlined.
Performance evaluation with AUC (A), precision (P), and recall (R) (with the standard mean error in brackets)
| Comparison . | Method . | EW-T2D . | LC . | C-T2D . | IBD . | Obesity . | |
|---|---|---|---|---|---|---|---|
| A. Taxonomic features | RF | A | 78.2%(3.2%) | 95.1%(1.5%) | 75.1%(2.6%) | 92.6%(2.2%) | 65.3%(4.4%) |
| P | 71.3%(2.3%) | 92.6%(1.7%) | 66.9%(1.9%) | 70.0%(20.0%) | 64.1%(0.4%) | ||
| R | 80.0%(3.4%) | 82.5%(3.6%) | 62.9%(3.9%) | 20.0%(6.3%) | 97.6%(1.5%) | ||
| SVM | A | 75.5%(3.1%) | 89.4%(2.2%) | 71.6%(2.4%) | 92.0%(2.6%) | 64.2%(3.8%) | |
| P | 75.2%(3.6%) | 81.9%(3.3%) | 65.7%(3.5%) | 66.7%(18.3%) | 70.9%(2.5%) | ||
| R | 69.1%(6.2%) | 75.0%(4.2%) | 61.8%(4.7%) | 36.0%(9.8%) | 77.0%(3.1%) | ||
| MLP | A | 74.1%(5.1%) | 90.0%(1.5%) | 72.3%(2.2%) | 86.4%(5.0%) | 52.3%(3.0%) | |
| P | 86.4%(6.9%) | 79.3%(2.1%) | 67.3%(2.0%) | 86.3%(2.3%) | 67.2%(1.5%) | ||
| R | 69.1%(6.8%) | 86.1%(3.2%) | 64.1%(5.8%) | 94.1%(2.6%) | 70.3%(1.5%) | ||
| FT-Transformer | A | 78.4%(3.6%) | 90.4%(2.4%) | 73.2%(3.0%) | 93.6%(2.0%) | 66.1%(1.7%) | |
| P | 71.4%(4.7%) | 84.8%(6.2%) | 66.8%(3.2%) | 70.1%(7.7%) | 75.9%(2.5%) | ||
| R | 69.1%(6.8%) | 85.8%(3.4%) | 64.1%(3.3%) | 72.0%(4.9%) | 67.3%(6.9%) | ||
| DeepMicro [7] | A | 82.9%(3.9%) | 88.8%(1.1%) | 72.5%(2.5%) | 87.3%(3.0%) | 67.4%(3.4%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| EnsDeepDP [4] | A | 86.7%(N/A) | 94.3%(N/A) | 77.6%(N/A) | 95.8%(N/A) | 72.3%(N/A) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| B. Taxonomic features with external knowledge | Meta-Signer [15] | A | – | 90.5%(5.0%) | – | 79.4(15.9%) | 60.0%(13.5%) |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| TaxoNN [16] | A | – | 91.1%(N/A) | 73.3%(N/A) | – | – | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| PopPhy-CNN [14] | A | – | 90.1%(N/A) | – | – | 58.9%(N/A) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| MicroKPNN [18] | A | 85.8(6.7%) | 96.9%(0.9%) | 75.5%(3.2%) | 95.4%(3.7%) | 72.8%(4.8%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| C. Taxonomic features and functional features | MDL4Microbiome | A | 76.8%(5.7%) | 92.5%(1.3%) | 74.7%(1.3%) | 88.0%(0.6%) | 54.1%(2.0%) |
| P | 76.3%(7.5%) | 78.1%(1.4%) | 67.6%(1.9%) | 86.0%(1.7%) | 69.3%(1.9%) | ||
| R | 69.1%(4.6%) | 89.1%(5.4%) | 66.5%(4.2%) | 92.9%(2.9%) | 72.1%(3.2%) | ||
| MVIB [29] | A | 85.9%(2.3%) | 92.5%(0.5%) | 75.8%(1.2%) | 93.6%(1.4%) | 66.6%(2.7%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| FT-Concat | A | 82.6%(3.1%) | 91.7%(1.7%) | 77.5%(1.9%) | 92.0%(2.4%) | 61.5%(4.2%) | |
| P | 77.8%(4.0%) | 89.4%(3.2%) | 68.9%(2.2%) | 79.3%(9.7%) | 75.1%(3.3%) | ||
| R | 80.0%(7.3%) | 76.7%(4.5%) | 68.2%(6.1%) | 64.0%(7.5%) | 57.6%(12.4%) | ||
| FT-Vote | A | 82.2%(2.1%) | 91.4%(1.4%) | 76.6%(1.6%) | 94.6%(1.7%) | 61.0%(3.3%) | |
| P | 70.5%(5.4%) | 84.4%(4.0%) | 68.3%(1.5%) | 77.7%(6.1%) | 70.4%(2.1%) | ||
| R | 80.0%(6.0%) | 87.5%(2.9%) | 63.5%(4.3%) | 56.0%(11.7%) | 63.0%(11.9%) | ||
| T-MBT | A | 83.2%(4.2%) | 91.7%(2.3%) | 76.6%(1.6%) | 95.5%(2.1%) | 62.2%(3.3%) | |
| P | 80.0%(2.8%) | 87.1%(4.2%) | 69.7%(2.8%) | 81.7%(7.6%) | 73.1%(2.6%) | ||
| R | 67.3%(6.2%) | 80.8%(6.0%) | 70.0%(1.7%) | 68.0%(8.0%) | 60.6%(7.3%) | ||
| MSFT-Transformer | A | 87.5%(4.1%) | 94.1%(1.2%) | 77.6%(1.6%) | 97.4%(1.8%) | 67.7%(2.9%) | |
| P | 80.5%(7.0%) | 88.7%(3.9%) | 70.8%(2.8%) | 100.0%(0.0%) | 72.7%(2.9%) | ||
| R | 83.6%(6.0%) | 89.2%(3.4%) | 65.3%(2.5%) | 72.0%(13.6%) | 69.7%(10.5%) | ||
| Comparison . | Method . | EW-T2D . | LC . | C-T2D . | IBD . | Obesity . | |
|---|---|---|---|---|---|---|---|
| A. Taxonomic features | RF | A | 78.2%(3.2%) | 95.1%(1.5%) | 75.1%(2.6%) | 92.6%(2.2%) | 65.3%(4.4%) |
| P | 71.3%(2.3%) | 92.6%(1.7%) | 66.9%(1.9%) | 70.0%(20.0%) | 64.1%(0.4%) | ||
| R | 80.0%(3.4%) | 82.5%(3.6%) | 62.9%(3.9%) | 20.0%(6.3%) | 97.6%(1.5%) | ||
| SVM | A | 75.5%(3.1%) | 89.4%(2.2%) | 71.6%(2.4%) | 92.0%(2.6%) | 64.2%(3.8%) | |
| P | 75.2%(3.6%) | 81.9%(3.3%) | 65.7%(3.5%) | 66.7%(18.3%) | 70.9%(2.5%) | ||
| R | 69.1%(6.2%) | 75.0%(4.2%) | 61.8%(4.7%) | 36.0%(9.8%) | 77.0%(3.1%) | ||
| MLP | A | 74.1%(5.1%) | 90.0%(1.5%) | 72.3%(2.2%) | 86.4%(5.0%) | 52.3%(3.0%) | |
| P | 86.4%(6.9%) | 79.3%(2.1%) | 67.3%(2.0%) | 86.3%(2.3%) | 67.2%(1.5%) | ||
| R | 69.1%(6.8%) | 86.1%(3.2%) | 64.1%(5.8%) | 94.1%(2.6%) | 70.3%(1.5%) | ||
| FT-Transformer | A | 78.4%(3.6%) | 90.4%(2.4%) | 73.2%(3.0%) | 93.6%(2.0%) | 66.1%(1.7%) | |
| P | 71.4%(4.7%) | 84.8%(6.2%) | 66.8%(3.2%) | 70.1%(7.7%) | 75.9%(2.5%) | ||
| R | 69.1%(6.8%) | 85.8%(3.4%) | 64.1%(3.3%) | 72.0%(4.9%) | 67.3%(6.9%) | ||
| DeepMicro [7] | A | 82.9%(3.9%) | 88.8%(1.1%) | 72.5%(2.5%) | 87.3%(3.0%) | 67.4%(3.4%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| EnsDeepDP [4] | A | 86.7%(N/A) | 94.3%(N/A) | 77.6%(N/A) | 95.8%(N/A) | 72.3%(N/A) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| B. Taxonomic features with external knowledge | Meta-Signer [15] | A | – | 90.5%(5.0%) | – | 79.4(15.9%) | 60.0%(13.5%) |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| TaxoNN [16] | A | – | 91.1%(N/A) | 73.3%(N/A) | – | – | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| PopPhy-CNN [14] | A | – | 90.1%(N/A) | – | – | 58.9%(N/A) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| MicroKPNN [18] | A | 85.8(6.7%) | 96.9%(0.9%) | 75.5%(3.2%) | 95.4%(3.7%) | 72.8%(4.8%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| C. Taxonomic features and functional features | MDL4Microbiome | A | 76.8%(5.7%) | 92.5%(1.3%) | 74.7%(1.3%) | 88.0%(0.6%) | 54.1%(2.0%) |
| P | 76.3%(7.5%) | 78.1%(1.4%) | 67.6%(1.9%) | 86.0%(1.7%) | 69.3%(1.9%) | ||
| R | 69.1%(4.6%) | 89.1%(5.4%) | 66.5%(4.2%) | 92.9%(2.9%) | 72.1%(3.2%) | ||
| MVIB [29] | A | 85.9%(2.3%) | 92.5%(0.5%) | 75.8%(1.2%) | 93.6%(1.4%) | 66.6%(2.7%) | |
| P | – | – | – | – | – | ||
| R | – | – | – | – | – | ||
| FT-Concat | A | 82.6%(3.1%) | 91.7%(1.7%) | 77.5%(1.9%) | 92.0%(2.4%) | 61.5%(4.2%) | |
| P | 77.8%(4.0%) | 89.4%(3.2%) | 68.9%(2.2%) | 79.3%(9.7%) | 75.1%(3.3%) | ||
| R | 80.0%(7.3%) | 76.7%(4.5%) | 68.2%(6.1%) | 64.0%(7.5%) | 57.6%(12.4%) | ||
| FT-Vote | A | 82.2%(2.1%) | 91.4%(1.4%) | 76.6%(1.6%) | 94.6%(1.7%) | 61.0%(3.3%) | |
| P | 70.5%(5.4%) | 84.4%(4.0%) | 68.3%(1.5%) | 77.7%(6.1%) | 70.4%(2.1%) | ||
| R | 80.0%(6.0%) | 87.5%(2.9%) | 63.5%(4.3%) | 56.0%(11.7%) | 63.0%(11.9%) | ||
| T-MBT | A | 83.2%(4.2%) | 91.7%(2.3%) | 76.6%(1.6%) | 95.5%(2.1%) | 62.2%(3.3%) | |
| P | 80.0%(2.8%) | 87.1%(4.2%) | 69.7%(2.8%) | 81.7%(7.6%) | 73.1%(2.6%) | ||
| R | 67.3%(6.2%) | 80.8%(6.0%) | 70.0%(1.7%) | 68.0%(8.0%) | 60.6%(7.3%) | ||
| MSFT-Transformer | A | 87.5%(4.1%) | 94.1%(1.2%) | 77.6%(1.6%) | 97.4%(1.8%) | 67.7%(2.9%) | |
| P | 80.5%(7.0%) | 88.7%(3.9%) | 70.8%(2.8%) | 100.0%(0.0%) | 72.7%(2.9%) | ||
| R | 83.6%(6.0%) | 89.2%(3.4%) | 65.3%(2.5%) | 72.0%(13.6%) | 69.7%(10.5%) | ||
The reported results of cited methods are based on previous works that use the same datasets. There are a total of nine methods (uncited) evaluated with our framework for five trails, including MSFT-Transformer. N/A indicates that the corresponding study did not provide the standard mean error. The best performance of each dataset is highlighted in bold while the second-high underlined.
This PDF is available to Subscribers Only
View Article Abstract & Purchase OptionsFor full access to this pdf, sign in to an existing account, or purchase an annual subscription.
