Dynamic and Static Features-Aware Recommendation with Graph Neural Networks

Table 4

Performance of compared with different variants of DSAGR (“—” indicates DSAGR removes the key technology).

Variants	Dynamic preference		Dynamic feature	ML-100k		ML_1M
	Long-term	Short-term		Recall@N	NDCG@N	Recall@N	NDCG@N
DSAGR-S	—	GCN	CNN	0.36013	0.44148	0.2891	0.3214
DSAGR-L	GCN	—	CNN	0.36370	0.32779	0.3000	0.3325
DSAGR-G	GCN	—	—	0.33217	0.4041	0.2345	0.2911
DSAGR-D	GCN	GCN	—	0.36006	0.44097	0.2798	0.3154
DSAGR-DL	GCN	GCN	LSTM	0.36184	0.44274	0.2921	0.3255
DSAGR-DG	GCN	GCN	GRU	0.36456	0.44921	0.2966	0.3310
DSAGR	GCN	GCN	CNN	0.3672	0.4470	0.3054	0.3421