Recommendation Based on Users’ Long-Term and Short-Term Interests with Attention

Table 2

Performance comparison in terms of Recall@20 and NDCG on all datasets.

Dataset	Metric	S-POP	BPRMF	FPMC	Item-KNN	GRURec	GRU4RecCon	HRNN Init	Att-BNN
ML-100K	Recall@20	0.3044	0.3131	0.4562	0.5160	0.6064	0.6784	0.6935	0.7094
	NDCG	0.1835	0.1208	0.1501	0.2108	0.2289	0.2900	0.3087	0.4108
ML-1M	Recall@20	0.2708	0.3400	0.4125	0.5231	0.5953	0.6911	0.7031	0.7344
	NDCG	0.1775	0.0157	0.1619	0.2170	0.2260	0.2922	0.3047	0.4258
Android app	Recall@20	0.2107	0.1519	0.3155	0.2835	0.4382	0.5795	0.5821	0.6391
	NDCG	0.1469	0.0863	0.0892	0.0945	0.1546	0.2493	0.3047	0.5022
Digital music	Recall@20	0.1206	0.1321	0.2107	0.1680	0.2896	0.4872	0.4966	0.6231
	NDCG	0.0773	0.0731	0.0820	0.0822	0.1883	0.2043	0.3152	0.4671
Instant video	Recall@20	0.2315	0.1724	0.3428	0.3147	0.4526	0.6031	0.6119	0.6359
	NDCG	0.1257	0.0913	0.0936	0.0917	0.1629	0.2503	0.2634	0.4062
Last FM	Recall@20	0.3162	0.2752	0.3769	0.4265	0.5371	0.7254	0.8105	0.8735
	NDCG	0.2355	0.0861	0.1048	0.1327	0.1852	0.3372	0.4759	0.5582
JD	Recall@20	0.5854	0.1873	0.2433	0.1246	0.7034	0.7934	0.8073	0.8593
	NDCG	0.2176	0.0664	0.0496	0.0361	0.4198	0.5932	0.6098	0.6742