Table 1.
Individual and ensemble methods prediction performances on MIND test set.
| Algorithm | MRR | Hits@1 | Hits@3 | Hits@10 |
|---|---|---|---|---|
|
| ||||
| CBR | 0.0128 | 0.0020 | 0.0156 | 0.0352 |
| probCBR | 0.2557 | 0.1816 | 0.3008 | 0.3984 |
| Rephetio | 0.0816 | 0.0488 | 0.0781 | 0.1445 |
| TransE | 0.1601 | 0.0645 | 0.1816 | 0.3770 |
| DistMult | 0.0391 | 0.0098 | 0.0293 | 0.0781 |
| ComplEx | 0.0820 | 0.0273 | 0.0801 | 0.2031 |
| RotatE | 0.1396 | 0.0840 | 0.1406 | 0.2637 |
| Intersection (7) | 0.9792 | 0.9583 | 1.000 | 1.000 |
| Union (7) | 0.1613 | 0.0898 | 0.1602 | 0.3379 |
Intersection & Union (7) represents MRR and ‘hits at k’ calculated with seven algorithms and their respective policies. CBR, probCBR, and Rephetio are path-based approaches. TransE, DistMult, ComplEx, and RotatE are embedding-based approaches.