The codes of MSTAN are implemented and tested under the following development environment:
- python=3.7.13
- tensorflow-gpu=2.5.0
- numpy=1.19.5
We used two datasets, Foursquare and Gowalla, as mentioned in our paper.
The datasets are divided into training set, validation set and testing set by 8:1:1.
We provided the Foursquare dataset in data/Foursquare.
You can run the code from the following command line:
python main.py
In addition, you can debug hyperparameters in code.