PeriodicMove: Shift-aware Human Mobility Recovery with Graph Neural Network

This is the source code for “PeriodicMove: Shift-aware Human Mobility Recovery with Graph Neural Network” accepted by CIKM 2021:

Dependencies

• Python==3.6
• torch==1.5.0
• transformers==2.9.0
• easydict==1.9
• matplotlib==3.1.1

Running

For the simplification, we only supply the foursquare dataset used in the paper.

Foursquare_mask_num_10.tar.gz

Firstly, the user should extract the file, e.g., pos.vocab.txt. Then they need to change the setting in the config.

vocab_path      # the path of vocab file
dist_path       # the path of distance file
train_file      # the path of traning data
eval_path       # the path of validation data
test_path       # the path of testing data
save_dir        # the path for saving model and embedding matrix

After that, type the following command in the termination.

python main.py

When the training procedure is completed, the terminal will print the results stated in our paper.