A Novel End-to-end Framework for A-share Stock Market Portfolio Optimization Considering Risk Measure and Feature Exposure
Year: 2024
Type: conference-paper
Abstract: In the current financial markets, where portfolio optimization remains a significant challenge, the dynamic changes in market conditions and investment demands render the development of an effective portfolio optimization framework more crucial than ever. This study introduces an end-to-end portfolio optimization framework, termed Quality-Growth-Momentum-Sentiment-Bayesian Optimization Asset Portfolio (QBOAP), aimed at constructing robust and profitable asset portfolios within the A-share stock market. The framework enhances both portfolio performance and trading value. The QBOAP framework utilizes the Quality-Growth-Momentum-Sentiment (QGMS) feature set to capture a comprehensive array of market information, ensuring that feature combinations are both representative and explanatory. A novel loss function is employed within this framework, incorporating considerations of return curve stability and feature exposure, and is optimized using a Bayesian optimization algorithm to determine the optimal asset weights. Additionally, the framework integrates a weight rank-based trading strategy designed to minimize transaction costs and complexity, thereby improving overall trading value. Empirical validation through backtesting from 2017 to 2024 reveals that the QBOAP framework achieves a Compound Annual Growth Rate (CAGR) of 0.216, a Sharpe ratio of 1.46, and a Calmar ratio of 1.60. These metrics are superior to those of ten baseline models, demonstrating the framework's effectiveness and offering valuable insights for the construction and optimization of A-share portfolios.
Authors: Nanxi Xu, Haotian Xiao, Yukun Zhu, Xiaochi Chen, Yixuan Li, Xiaoli Hu
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: bronze
Funder: Jianghan University
Award: G8361316409
OpenAlex ID: W4405398528