Leveraging Multiactions to Improve Medical Personalized Ranking for Collaborative Filtering

Table 3

Recommendation performance of different methods on the dataset from Topmd and row “Improve” shows the percentage of improvements that MBPR achieves relative to the best baseline method.

	Pre@5	Recall@5	AUC	MAP	NDCG	MRR
Random	0.0010	0.0042	0.4989	0.0070	0.1248	0.0076
BPR-MF	0.0154	0.0720	0.8384	0.0577	0.2043	0.0602
MBPR-1	0.0154	0.0723	0.8427	0.0584	0.2051	0.0610
MBPR-2	0.0155	0.0725	0.8427	0.0580	0.2048	0.0606
Improve	0.64%	0.69%	0.51%	1.21%	0.39%	1.32%
BPR-MF	0.0160	0.0749	0.8383	0.0587	0.2052	0.0614
MBPR-1	0.0172	0.0801	0.8304	0.0672	0.2123	0.0707
MBPR-2	0.0172	0.0800	0.8388	0.0629	0.2095	0.0661
Improve	7.50%	6.94%	0.05%	14.48%	3.46%	15.14%