Dynamic and Static Features-Aware Recommendation with Graph Neural Networks

Table 5

Performance of 16 experiments obtained from Taguchi’s method.

ID	Time slices	Filter-first CNN layer	Filter-second CNN layer	Embedding factors	Recall@N	NDCG@N
1	11	2	8	8	0.3422	0.4200
2	11	4	16	16	0.3552	0.4369
3	11	6	32	32	0.3617	0.4420
4	11	8	64	64	0.3540	0.4296
5	21	2	16	32	0.3560	0.4373
6	21	4	8	64	0.3626	0.4412
7	21	6	64	8	0.3451	0.4224
8	21	8	32	16	0.3438	0.4186
9	31	2	32	64	0.3672	0.4470
10	31	4	64	32	0.3660	0.4481
11	31	6	8	16	0.3463	0.4237
12	31	8	16	8	0.3469	0.4259
13	41	2	64	16	0.3547	0.4232
14	41	4	32	8	0.3483	0.41786
15	41	6	16	64	0.3533	0.4217
16	41	8	8	32	0.3491	0.4286