ActSTD

The official implementation of: Activity Trajectory Generation via Modeling Spatiotemporal Dynamics (KDD '22).

Installation

Environment

• Tested OS: Linux
• Python >= 3.7
• PyTorch == 1.7.1

Dependencies

1. Install PyTorch 1.7.1 with the correct CUDA version.
2. Use the pip install -r requirements. txt command to install all of the Python modules and packages used in this project.

run python src/setup.py build_ext --inplace to create the shared object file in the current directory.

Model Training

cd src

Use the following command to train ActSTD on Mobile dataset with different CNF models:

python app.py --data Mobile --model attncnf --tpp neural  --l2_attn --ode_method 'scipy_solver' --ode_solver 'RK45' --cuda_id 0 --tpp_style 'gru' --weekhour

python app.py --data Mobile --model jumpcnf --tpp neural --solve_reverse --ode_method 'scipy_solver' --ode_solver 'RK45' --cuda_id 0 --tpp_style 'gru' --weekhour

Use the following command to train ActSTD on Foursquare dataset with different CNF models:

python app.py --data Foursquare --model attncnf --tpp neural  --l2_attn --ode_method 'scipy_solver' --ode_solver 'RK45' --cuda_id 0 --tpp_style 'gru' --weekhour

python app.py --data Foursquare --model jumpcnf --tpp neural --solve_reverse --ode_method 'scipy_solver' --ode_solver 'RK45' --cuda_id 0 --tpp_style 'gru' --weekhour

More Related Works

• Learning to Simulate Daily Activities via Modeling Dynamic Human Needs (WWW'23)

Citation

If you find this repository useful in your research, please consider citing the following paper:

@inproceedings{yuan2022activity,
  title={Activity Trajectory Generation via Modeling Spatiotemporal Dynamics},
  author={Yuan, Yuan and Ding, Jingtao and Wang, Huandong and Jin, Depeng and Li, Yong},
  booktitle={Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
  pages={4752--4762},
  year={2022}
}

Note

The implemention is based on Neural STPP.