The code for paper: Biased Temporal Convolution Graph Network for Time Series Forecasting with Missing Values.

1. Install requirements. pip install -r requirements.txt
2. Download data.

Download Metr-LA, ETTh1, Electricity, PEMS datasets from here. Obtain BeijingAir dataset from Brits. Put all the files in the ./data.

3. Training.

python main.py --epochs 200 --mask_ratio 0.2 --dataset-name Metr

4. Testing.

python test_forecasting.py --epochs 200 --mask_ratio 0.2 --dataset-name Metr

@inproceedings{BiTGraph, 
  title={Biased Temporal Convolution Graph Network for Time Series Forecasting with Missing Values},
  author={Chen, Xiaodan and Li, Xiucheng and Liu, Bo and Li, Zhijun},
  booktitle={International Conference on Learning Representations (ICLR)},
  year={2024}
}