This codebase contains the implementation for the paper An asset subset-constrained minimax optimization framework for online portfolio selection (ESWA).
In this paper, we propose an asset subset-constrained minimax (ASCM) optimization framework, which generates optimal portfolios from diverse investment strategies represented as asset subsets.
- Operating System: tested on Window 10.
- Python Version: 3.8.0.
You can install the dependencies by:
pip install -r requirements.txt
We tested the algorithms performance on six real datasets, namely djia, hs300, crypto, sp500, tse, and EuroStoxx50 which stored in universal\data directory. The first row in the dataset represents the stock name, and each subsequent row represents the closing price of all stocks for the same trading period. If you want to test other datasets, you can first store the dataset in pkl format in the data directory, and then modify the list of test datasets in the testSimple() function under SimpleTest.py.
This project includes two tested algorithms, namely ASCM and benchmark algorithm BAH(Buy ang Hold), which are stored in universal\algos directory. If you want to add other comparison algorithms, you can emulate the ascm algorithm, rewrite the step() function, store the [algorithm name].py file in the universal\algos directory, import the algorithm package in __init__.py in that directory, and finally add the algorithm running code in the showResult() function in the SimpleTest.py.
The hyperparameters of each algorithm and transaction cost can be modified according to your needs in the function showResult() under SimpleTest.py. For instance, there is an algorithm instance result_ascm, and you can modify its transaction cost to 1‰:
result_ascm.fee = 0.001
You can run the code using the following command:
python SimpleTest.py
By default, the program will output the prices of the current test dataset, the specific progress of each algorithm, various performance indicators (cumulative wealth, sharpe ratio, max drawdown, etc.) of ASCM, and an EPS format image which will be stored in experimentFigure\cumulative wealth\[dataset name].eps.
If you find our paper or code is useful, please consider citing:
@article{yin2024asset,
title={An asset subset-constrained minimax optimization framework for online portfolio selection},
author={Yin, Jianfei and Zhong, Anyang and Xiao, Xiaomian and Wang, Ruili and Huang, Joshua Zhexue},
journal={Expert Systems with Applications},
pages={124299},
volume = {254},
year={2024},
publisher={Elsevier}
}