source_code for TULVAE (implement of TULVAE model).
--Learning Hierarchical Structures and Latent Representation of Human Mobility for Trajectory-User Linking
Tensorflow 1.0 or++
python 2.7
numpy
We have shown the details of trajectory splitting in the paper. For embbedding, you can choose the wordv2vec toolkit.
To run TULVAE, run the following commands:
'python TULVAE_1122.py'
In each data, for example (Gowalla):
0 480992 49904 420315 73407 (a sample line in our dataset)
The first column is the user id, other columns are a list of POIs generated within 6 hours.
We use four different Location-based Social Network datasets as follows.
- Gowalla: http://snap.stanford.edu/data/loc-gowalla.html
- Brightkite: http://snap.stanford.edu/data/loc-brightkite.html
- Foursquare(New York): https://sites.google.com/site/yangdingqi/home/foursquare-dataset
- (remark) Please do not use these datasets for commercial purpose. For academic uses, please refer to the paper.
Any comments and feedback are appreciated.