Code repository for this paper:
StockFormer: Learning Hybrid Trading Machines with Predictive Coding.
Siyu Gao, Yunbo Wang^†, Xiaokang Yang

git clone https://github.com/gsyyysg/StockFormer.git
cd StockFormer
pip install -r requirements.txt

Downloaded from YahooFinance

dir: 'data/CSI/'

dir:'code/'

1）Relational state inference module training:

cd code/Transformer/script
sh train_mae.sh

2）Long-term state inference module training:

cd code/Transformer/script
sh train_pred_long.sh

3. Short-term state inference module training:

cd code/Transformer/script
sh train_pred_short.sh

4. Select the best model of three state inference modules from 'code/Transformer/checkpoints/' according to their performance on validation set and add them to 'code/Transformer/pretrained/'

OR directly use the model which have been pretrained in advance by us (dir:'code/Transformer/pretrained/csi/ ')

1. train SAC model (three state inference module's path can be changed in train_rl.py file)

python train_rl.py

2. get prediction result on test set from 'code/results/df_print/'

If you find our work helps, please cite our paper.

@inproceedings{gaostockformer,
  title={StockFormer: Learning Hybrid Trading Machines with Predictive Coding},
  author={Gao, Siyu and Wang, Yunbo and Yang, Xiaokang},
  booktitle={IJCAI},
  year={2023}
}

This codebase is based on FinRL.