G4P

Code for the paper Portfolio Selection with Graph-aware Gaussian Process Regression

Prerequests

our implementation is mainly based on following packages:

python 3.7
pip install gpuinfo
pip install universal-portfolios
pip install tensorflow-gpu==1.15
pip install gpflow==1.5

Besides, some basic packages like numpy are also needed.

Take the NYSE (N) dataset for example.

Name		Name	Last commit message	Last commit date
compatible		compatible
dgp_graph		dgp_graph
doubly_stochastic_dgp		doubly_stochastic_dgp
gpflow_monitor		gpflow_monitor
jack_utils		jack_utils
stock_data		stock_data
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
baselines-synthetic-fee0.ipynb		baselines-synthetic-fee0.ipynb
baselines.ipynb		baselines.ipynb
err_analysis.ipynb		err_analysis.ipynb
main.py		main.py
metrics.py		metrics.py
preprocess_MSCI.ipynb		preprocess_MSCI.ipynb
preprocess_NYSE.ipynb		preprocess_NYSE.ipynb
preprocess_NYSE_N.ipynb		preprocess_NYSE_N.ipynb
preprocess_SNP500.py		preprocess_SNP500.py
preprocess_TSE.ipynb		preprocess_TSE.ipynb
run_MSCI.sh		run_MSCI.sh
run_NYSE.sh		run_NYSE.sh
run_NYSE_N.sh		run_NYSE_N.sh
run_SNP500.sh		run_SNP500.sh
run_TSE.sh		run_TSE.sh
run_synthetic.sh		run_synthetic.sh
strategies.py		strategies.py
strategy-MSCI.ipynb		strategy-MSCI.ipynb
strategy-NYSE.ipynb		strategy-NYSE.ipynb
strategy-NYSE_N.ipynb		strategy-NYSE_N.ipynb
strategy-SNP500-sensitivity.ipynb		strategy-SNP500-sensitivity.ipynb
strategy-SNP500.ipynb		strategy-SNP500.ipynb
strategy-TSE.ipynb		strategy-TSE.ipynb
strategy-synthetic-fee0.ipynb		strategy-synthetic-fee0.ipynb
utils.py		utils.py