Our dataset and codes are released from the paper.

Geographcial part of our model VGMF is located at /experiment/model/geo, where KMeansGMF.py is the variant GMF. Visual part is located at /experiment/model/images, where /train/MF_visual.py is the variant VMF and /train/KMeansVGMF.py is the final model VGMF.

It includes two parts, .csv files are check-in records of users, which are divided into train, validate and test dataset, and .tfrecords files are visual features of images posted by users and locations, which are generated after \experiment\model\images\train\extract_feartures.py running.

Please give citation when using our dataset or code in your own work:

@article{liuvgmf,  
  title={VGMF: Visual contents and geographical influence enhanced point-of-interest recommendation in location-based social network},  
  author={Liu, Bo and Meng, Qing and Zhang, Hengyuan and Xu, Kun and Cao, Jiuxin},  
  journal={Transactions on Emerging Telecommunications Technologies},  
  pages={e3889},  
  publisher={Wiley Online Library}