1. Environment Configuration

numpy==1.19.5 
matplotlib==3.3.4
pandas==1.1.5
scikit-learn==0.24.2 
scipy==1.5.4   
einops==0.4.1
torch==1.10.0
torchaudio==0.10.0
torchvision==0.11.0
cudatoolkit==11.3.1
gpu=NVIDIA Tesla P100-SXM2 GPU
gpu memory=16384 MiB
cpu=Intel(R) Xeon(R) Silver 4216 CPU @ 2.10GHz

1. Create a virtual environment: conda create -n pytorch_cuda python=3.6
2. Activate the environment: conda activate pytorch_cuda
3. Install the gpu version of Pytorch: conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3
4. Install other required packages.

2. Data

The folder ./dataset/ holds the 8 benchmark datasets used in the experiment.

3. Code Download

git clone https://github.com/linxi20/CPAT.git

4. Experimental Parameters

The parameter settings for all datasets are stored in the sh script file under the folder ./scripts/.

5. Model Training and Evaluation

1、We provide the running scripts for all benchmark tests under the folder ./scripts/. You can reproduce the experiments results by the following example:

• Switch to the folder CPAT/scripts:

   cd CPAT/scripts
  

• Submit all the script files below:

   # Multivariate forecasting with CPAT
   nohup ./illness.sh > illness.log 2>&1 
   nohup ./etth1.sh > etth1.log 2>&1 
   nohup ./etth2.sh > etth2.log 2>&1 
   nohup ./ettm1.sh > ettm1.log 2>&1 
   nohup ./ettm2.sh > ettm2.log 2>&1 
   nohup ./weather.sh > weather.log 2>&1 
   nohup ./electricity.sh > electricity.log 2>&1 
   nohup ./traffic.sh > traffic.log 2>&1 
  

2、After the scripts runs, the folder ./logs/LongForecasting/ is created in the current directory, which stores the log files recording the training process. The experimental results can be viewed via result.txt after the training is completed and the prediction accuracy of the model can be evaluated based on the MSE and MAE metrics.