This is an open-source project for deep learning researchers, especially for time series forecasting. The repo will release in the future. We provide a detailed code project to evaluate advanced deep time series models long-term forecasting.
🌟 Considering the time lag effect between the forecasting output and the input in time series forecasting, FPPformer proposes a new predictive attention to achieve goal. A predictive attention for time series forecasting.
- Install Python 3.8 or later. For convenience, execute the following command.
pip install -r requirements.txt
-
Prepare Data. You can obtain the all datasets from folder
./data. -
Train and evaluate model. We provide the experiment scripts for all benchmarks under the folder
./scripts/. You can reproduce the experiment results as the following examples:
# long-term forecast
bash ./scripts/ETT_script/FPPformer_ETTm2.sh
bash ./scripts/ECL_script/FPPformer_ECL.sh
bash ./scripts/Exchange_script/FPPformer_Exchange.sh
bash ./scripts/PEMS_script/FPPformer_PEMS.sh
bash ./scripts/Traffic_script/FPPformer_Traffic.sh
bash ./scripts/Weather_script/FPPformer_WTH.sh
- Develop your own model.
- Add the model file to the folder
./models. You can follow the./models/FPPformer.py. - Include the newly added model in the
Exp_Basic.model_dictof./exp/exp_basic.py. - Create the corresponding scripts under the folder
./scripts.
If you have any questions or suggestions, feel free to contact:
- Chengli Zhou ([email protected])
- Chunna Zhao ([email protected])
Or describe it in Issues.
PFformer outperforms six other state-of-the-art models with just consider one state lag.
