A Joint Deep Recommendation Framework for Location-Based Social Networks

Table 1

Performance comparison over the Yelp dataset. Improvement of TCENR compared to each method is shown in brackets.

Model	Accuracy	MSE	Pre@10	Rec@10
HPF	0.8141	0.1800	0.5526	0.3699
	(1.69%)	(34.94%)	(18.51%)	(40.98%)
NMF	0.8222	0.1189	0.7851	0.3517
	(0.69%)	(1.51%)	(-16.58)	(48.28%)
Geo-SAGE	0.7995	0.1807	0.2912	0.4145
	(3.55%)	(35.19%)	(124.89%)	(25.81%)
LCARS	0.8142	0.1612	0.6408	0.5127
	(1.68%)	(27.36%)	(2.2%)	(1.72%)
NeuMF	0.8273	0.1421	0.6488	0.5586
	(0.07%)	(17.59%)	(0.94%)	(-6.64)
Pace	0.8239	0.1186	0.6406	0.5049
	(0.49%)	(1.26%)	(2.23%)	(3.29%)
DeepCoNN	0.8037	0.1454	0.5385	0.323
	(3.01%)	(19.46%)	(21.62%)	(64.46%)
TCENR	0.8279	0.1171	0.6549	0.5215
	0.8273	0.1161	0.6655	0.4738
	(0.07%)	(-0.86)	(-1.59)	(10.07%)