Adaptive Alleviation for Popularity Bias in Recommender Systems with Knowledge Graph

Table 2

Performance comparison of different recommendation models.

Dataset	Metric	BPRMF	LightGCN	KTUP	IPS-CN	ESAM	AWING	AWING-APS
LastFM (@10,%)	Recall	5.17	5.82	6.75	6.49	5.64	7.31	6.86
	NDCG	5.39	5.48	6.04	5.32	5.59	6.94	5.89
	GAP (N)	92.67	80.49	60.83	38.13	25.84	59.08	5.21
	GAP (D)	95.71	87.25	57.39	53.04	31.82	64.26	7.56
	GAP (B)	87.52	82.98	68.52	−60.49	−55.25	63.94	−4.81
	AD	5.62	12.07	17.61	16.47	18.15	16.24	24.28
	APT	1.56	3.76	14.79	23.59	25.33	14.66	36.06
DBbook (@10,%)	Recall	17.23	20.45	23.15	19.45	18.92	24.62	23.02
	NDCG	18.86	22.67	25.92	22.62	21.53	27.62	25.27
	GAP (N)	88.77	94.14	53.32	25.96	27.71	55.62	2.18
	GAP (D)	85.48	83.46	61.28	35.09	32.89	62.52	8.03
	GAP (B)	75.49	79.71	43.76	−34.28	−43.27	42.24	−5.33
	AD	9.58	11.26	14.36	17.33	17.61	15.38	19.78
	APT	2.91	5.51	7.95	12.06	13.17	12.27	17.68

The bold numbers indicate the best in each row and the underlined values indicate the second best.