ZhuZhouFan / CQVAE

This resposity is a pre-released verison of Python code used in the paper "Asset pricing via the conditional quantile variational autoencoder".

Asset pricing via the conditional quantile variational autoencoder

This resposity is a pre-released verison of Python code used in the paper "Asset pricing via the conditional quantile variational autoencoder".

If you have any question about this resposity or implement details of paper, please feel free to submit a new issue.

Reproduce the simulation results with certain hyperparameters

# CAE
python main.py --model CAE --save-folder logs/CAE --lam 1e-4

# CVAE
python main.py --model CVAE --save-folder logs/CVAE

# CQAE
python main.py --model CQAE --save-folder logs/CQAE --lam 2e-5

# CQVAE
python main.py --model CQVAE --save-folder logs/CQVAE

Reproduce the empirical analysis result

1. Download data from the homepage of Prof. Dacheng Xiu (homepage link and data link).
2. Preprocess the data by python preprocess_data.py. Please make sure the directories are specified.
3. Train the models by python empirical_main.py. The tuning procedure of hyperparameters and ensemble learning should be done in this step.
4. Calculate the out-of-sample prediction by using python inference_main.py.

Citation

Please cite our paper if you feel this code helps.

@article{Yang2024CQVAE,
    author = {Xuanling Yang, Zhoufan Zhu, Dong Li and Ke Zhu},
    title = {Asset Pricing via the Conditional Quantile Variational Autoencoder},
    journal = {Journal of Business \& Economic Statistics},
    volume = {42},
    number = {2},
    pages = {681--694},
    year = {2024},
    publisher = {Taylor \& Francis},
    doi = {10.1080/07350015.2023.2223683},
}