An informal implemantation of MoveSim

Codes for paper in KDD'20 (AI for COVID-19): Learning to Simulate Human Mobility

Modification

This code is based on the implementation of the authors (https://github.com/FIBLAB/MoveSim), but there are some modifications.

• Remove p_loss and d_loss because our using data does not satisfy the assumption of them.
• Replace the two self-attention network of the generator into one self-attention network because the network does not work well in our dataset.

Usage

To run this code using the test data

./run.sh

Citation

Feng, Jie, et al. "Learning to simulate human mobility." Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining. 2020.