MulSTE

This is the origin Pytorch implementation of MulSTE in the following paper:

[KDD' 24] MulSTE: A Multi-view Spatio-temporal Learning Framework with Heterogeneous Event Fusion for Demand-supply Prediction

Requirements

• Python 3.8
• matplotlib == 3.1.1
• numpy == 1.19.4
• pandas == 1.3.3
• torch == 1.9.0
• transformers == 4.11.3
• tqdm == 4.62.3
• lxml == 4.6.3

Dependencies can be installed using the following command:

pip install -r requirements.txt

Folder Structure

└── MulSTE
    ├── config                 
    │   |──  data.conf          
    │   |──  data_zz.conf                          
    ├── lib         
    │   |──  figure_plot.py 
    │   |──  utils.py
    │   |──  utils_fix_normalization.py
    │   |──  utils_fix_normalization_zz.py                       
    ├── model
    │   |──  fine_tuned_bert
    │   |──  pre_trained_bert
    │   |──  MulSTE_model.py
    │   |──  MulSTE_model_zz.py
    ├── MulSTE_train.py
    ├── MulSTE_train_zz.py 
    ├── README.md
    └── requirements.txt

Reproducibility

1. Acquire Pre-trained BERT: Download pytorch_model.bin from Hugging Face and put it into model\pre_trained_bert\chinese-macbert-base folder.
2. Fine-tune BERT: Run Model_Fine_Tuning.ipynb in model\fine_tuned_bert folder to fine-tune the Pre-trained BERT, and put the output fine-tuned-bert.model into model\fine_tuned_bert folder.
3. Demand-supply Prediction: Then, all training and inference processes can be run at once, as follows.

Shanghai:

python MulSTE_train.py

Zhengzhou:

python MulSTE_train_zz.py