Hyperparameter ranges for all considered baseline models

Algorithms topK shrink Similarity Normalize α β ℓ2 norm
kNN algorithms [5 to 1000] [0 to 1000] [cosine, jaccard, asymmetric, dice, tversky] [True, False] — — —
Graph based algorithms [5 to 1000] — — — [0 to 2], strategy: uniform distribution [0 to 2], strategy: uniform distribution —
EASER — — — — — — [1 to 1e7], strategy: log-uniform distribution

Optimal hyperparameter values

Algorithms topK shrink Similarity Normalize α β ℓ2 norm
ItemkNN 508 1000 cosine True — — —
UserkNN 146 1000 cosine True — — —
P3α 777 — — False 1.087096950563704 — —
RP3β 777 — — True 0.5663562161452378 0.001085447926739258 —
EASER — — — — — — 1e3

Statistics of the provided train-test splits

Users Items Interactions Density Data leakage Interactions per usermin Interactions per itemmin Interactions per usermax Interactions per itemmax Testing users
114,737 30,040 1,781,093 0.00052 No 10 5 393 3902 114,737

Full result table

Algorithms R@10 R@20 N@10 N@20 T-Time (s) P-Time (s) Average P-Time Per User (ms)
Random 0.000 0.001 0.000 0.000 0.000 255.800 2.229
TopPop 0.017 0.031 0.012 0.017 0.028 230.004 2.005
ItemkNN 0.084 0.123 0.066 0.080 157.444 277.501 2.419
UserkNN 0.057 0.084 0.046 0.055 680.042 227.959 1.987
P3α 0.049 0.069 0.040 0.047 35.863 232.069 2.023
RP3β 0.076 0.110 0.061 0.073 35.743 235.756 2.055
EASE 0.080 0.120 0.062 0.077 552.271 292.651 2.551
KGIN 0.058 0.097 0.040 0.054 117360.426 362.043 3.155

Optimal hyperparameter values

Algorithms topK shrink Similarity Normalize α β ℓ2 norm
ItemkNN 144 1000 cosine True — — —
UserkNN 83 258 tversky True — — —
P3α 496 — — False 0.7681732734954694 — —
RP3β 350 — — True 0.7681732734954694 0.4181395996963926 —
EASER — — — — — — 1e3


Statistics of the provided train-test splits

Users Items Interactions Density Data leakage Interactions per usermin Interactions per itemmin Interactions per usermax Interactions per itemmax Testing users
23,566 48,123 1,544,131 0.00136 Yes, Effected users: 17290 2 1 2881 1445 23,566

Full result table

Algorithms R@10 R@20 N@10 N@20 T-Time (s) P-Time (s) Average P-Time Per User (ms)
Random 0.000 0.000 0.000 0.001 0.000 66.346 2.815
TopPop 0.012 0.022 0.017 0.020 0.025 55.528 2.356
ItemkNN 0.134 0.182 0.172 0.179 165.776 61.343 2.603
UserkNN 0.100 0.130 0.144 0.143 72.863 55.763 2.366
P3α 0.142 0.194 0.190 0.196 99.728 63.280 2.685
RP3β 0.140 0.189 0.180 0.187 89.168 64.697 2.745
EASE 0.135 0.185 0.178 0.184 2173.599 120.293 5.104
KGIN 0.137 0.214 0.154 0.177 300845.273 151.912 6.446

Optimal hyperparameter values

Algorithms topK shrink Similarity Normalize α β ℓ2 norm
ItemkNN 125 1000 cosine True — — —
UserkNN 454 1000 tversky True — — —
P3α 496 — — False 0.7681732734954694 — —
RP3β 350 — — True 0.7681732734954694 0.4181395996963926 —
EASER — — — — — — 1e3


Statistics of the provided train-test splits

Users Items Interactions Density Data leakage Interactions per usermin Interactions per itemmin Interactions per usermax Interactions per itemmax Testing users
70,679 24,915 846,434 0.00048 No 4 5 4475 1740 70,591

Full result table

Algorithms R@10 R@20 N@10 N@20 T-Time (s) P-Time (s) Average P-Time Per User (ms)
Random 0.001 0.002 0.000 0.001 0.000 145.591 2.060
TopPop 0.012 0.029 0.007 0.012 0.013 133.699 1.892
ItemkNN 0.121 0.164 0.084 0.097 74.114 132.712 1.878
UserkNN 0.116 0.160 0.078 0.092 264.982 137.988 1.952
P3α 0.118 0.166 0.079 0.094 34.924 134.767 1.907
RP3β 0.105 0.144 0.072 0.084 37.803 135.821 1.922
EASER 0.116 0.161 0.079 0.092 314.753 159.325 2.254
KGIN 0.089 0.143 0.048 0.065 &ndash &ndash &ndash