EnhanceNet: Plugin Neural Networks for Enhancing Correlated Time Series Forecasting

This is a PyTorch implementation of EnhanceNet in the following paper:
Razvan-Gabriel Cirstea, Tung Kieu, Chenjuan Guo, Bin Yang, Sinno Jialin Pan, EnhanceNet: Plugin Neural Networks for Enhancing Correlated Time Series Forecasting , ICDE 2021. This work is based on DCRNN and Graph WaveNet. Being familiar with those models is strongly recommended.

Requirements

• torch
• scipy>=0.19.0
• numpy>=1.12.1
• pandas>=0.19.2
• pyyaml
• statsmodels
• torch
• tables
• future

Dependency can be installed using the following command:

pip install -r requirements.txt

Data Preparation

The traffic data files for Los Angeles (METR-LA) can be found here.

Run the Model on METR-LA

For RNN variants there are 4 configuration files which can be found under rnn/data/model. Each configuration corresponds to rnn/grnn with and without the dynamic weights. To run any of the models follow the command below, in addition to add the Dynamic Adjacency Matrix Generation Network add the argument --adaptive_supports=1 at the end of the command.

python dcrnn_train.py --config_filename=data/model/data/rnn.yaml

For TCN variants run the following command:

# TCN
python train.py 
# GTCN
python train.py --gcn_bool=True

In addition for adding the dynamic weights add --temporal_memory=1.

Citation

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

@inproceedings{cirstea2021enhancenet,
  title={EnhanceNet: Plugin Neural Networks for Enhancing Correlated Time Series Forecasting},
  author={Cirstea, Razvan-Gabriel and Kieu, Tung and Guo, Chenjuan and Yang, Bin and Pan, Sinno Jialin},
  booktitle={2021 IEEE 37th International Conference on Data Engineering (ICDE)},
  pages={1739--1750},
  year={2021},
  organization={IEEE}
}