Recommendation on Amazon Electronics (Recall, NDCG, MAP, EO, PRU, PRI)
14.3RecallOur model
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Our modelTop-M=50, Ablation=w/o Edge2025.01 | 14.3 | 6.31 | 2.61 | 91.92 | 44.63 | 42.86 | |
| Our modelTop-M=50, variant=2/32025.01 | 14.04 | 6.24 | 2.64 | 92.94 | 30.81 | 15.95 | |
| DPNSTop-M=502025.01 | 13.94 | 6.73 | 2.72 | 96.24 | 52.22 | 51.08 | |
| Our modelTop-M=50, Ablation=w/o Cost2025.01 | 13.77 | 6.04 | 2.5 | 91.24 | 33.11 | 22.8 | |
| APDATop-M=502025.01 | 12.86 | 6.16 | 2.45 | 89.63 | 50.61 | 35.72 | |
| Our modelTop-M=50, variant=1/22025.01 | 12.54 | 5.89 | 2.52 | 93.14 | 34.47 | 16.75 | |
| r-AdjNormTop-M=502025.01 | 12.28 | 5.84 | 2.31 | 93.04 | 47.84 | 35.53 | |
| LightGCNTop-M=502025.01 | 12.16 | 5.75 | 2.23 | 93.44 | 46.79 | 33.91 | |
| RegTop-M=502025.01 | 11.78 | 5.52 | 2.12 | 93.55 | 45.7 | 33.34 | |
| w/o Edgetop-M=20, ablation=Edge classification removal2025.01 | 8.44 | 4.78 | 2.39 | 95.47 | 44.31 | 42.68 | |
| Our model (2/3)top-M=20, ratio=2/32025.01 | 8.37 | 4.77 | 2.43 | 96.48 | 30.81 | 15.95 | |
| DPNStop-M=202025.01 | 8.23 | 5.15 | 2.51 | 97.58 | 52.6 | 51.4 | |
| w/o Costtop-M=20, ablation=Cost sensitive learning removal2025.01 | 8.18 | 4.59 | 2.28 | 95.27 | 32.95 | 22.57 | |
| Our model (1/2)top-M=20, ratio=1/22025.01 | 7.88 | 4.51 | 2.19 | 97.1 | 33.96 | 16.16 | |
| APDAtop-M=202025.01 | 7.64 | 4.67 | 2.21 | 92.4 | 50.15 | 35.31 | |
| LightGCNtop-M=202025.01 | 7.21 | 4.38 | 2.05 | 96.18 | 46.96 | 34.09 | |
| r-AdjNormtop-M=202025.01 | 7.16 | 4.44 | 2.12 | 94.72 | 47.11 | 35.18 | |
| Regtop-M=202025.01 | 6.78 | 4.15 | 1.95 | 95.02 | 43.93 | 31.49 |