Nonindependent Session Recommendation Based on Ordinary Differential Equation

Table 5

Performance of our model compared to the baseline algorithms.

Dataset	Diginetica			Yoochoose 1/64			Last.fm			Retailrocket
	Recall@20	Recall@50	MRR@20	Recall@20	Recall@50	MRR@20	Recall@20	Recall@50	MRR@20	Recall@20	Recall@50	MRR@20
POP	10.91	—	0.23	6.71	—	1.65	3.40	—	3.40	—	—	—
S-POP	21.07	—	14.69	30.44	—	18.35	19.36	—	11.47	—	—	—
GRU-rec	43.82	—	15.46	60.64	—	22.89	31.46	—	1.57	51.23	—	19.96
SR-GNN	—	—	17.59	—	—	30.94	41.20	—	21.89	66.60	—	30.18
NARM	62.58	—	27.35	68.32	—	28.76	37.73	—	13.12	68.95	—	28.33
Sess-ODEnet	69.98 + 7.4%	74.11	30.07 + 2.72%	69.19 + 0.87%	70.13	29.38 − 1.56%	49.66 + 8.46%	51.09	18.13 − 3.76%	71.15 + 2.2%	71.83	29.31 − 0.87%