Leveraging Contextual Influence and User Preferences for Point-of-Interest Recommendation

AUTHORS: Dongjin Yu, Wenbo Wanyan, Dongjing Wang

If there is any comment, please mail to yudj@hdu.edu.cn.

Experimental results

Source code:
DataSet:



Table 1. Statistics of datasets

Foursquare Gowalla
number of Users 7,642 5,628
number of POIs 28,483 31,803
number of Check-ins 512,523 620,683
Sparsity 99.76% 99.65%


Fig 1. Precision@N on Foursquare and Gowalla

Las Vegas


Fig 2. Recall@N on Foursquare and Gowalla

Toronto


Fig 3. F1@N on Foursquare and Gowalla

Toronto


Fig 4. Comparison on Foursquare and Gowalla with different sparsity

Toronto


Fig 5. Effect of embedding dimension and context window size

Toronto


Fig 6. EEffect of hyper-parameter γ

Toronto