Design and Evaluation of Quantitative Investment Strategies Using Reinforcement Learning and Multi-Factor Models
Wing-Yee Lam, Ka-Yan Cheung, Tsz-Hin Lau, Man-Kit Leung, Ho-Lam Chan · Frontiers in Business and Finance · 2025
Not cached yet. This record was fetched live for this request and is stored after the lookup. Reopening the page should serve the cached copy.
Frontiers in Business and Finance · OpenAlex · article · 2025
Abstract
Traditional quantitative investment models often suffer from limited adaptability in volatile market environments. To overcome this constraint, this study proposes a reinforcement learning-based framework, RL-Quant, which integrates technical indicators, sentiment signals, and fundamental variables into a multi-factor state representation. The agent is trained using the Proximal Policy Optimization (PPO) algorithm, with a customized reward function incorporating dynamic risk control parameters to constrain maximum drawdown and return volatility. Empirical backtesting is conducted on the CSI 300 and S&P 500 indices from 2014 to 2023.
The proposed framework achieves an annualized return of 19.6%, a maximum drawdown of 9.2%, and a Sharpe ratio of 1.87, consistently outperforming benchmark ETFs and equal-weighted portfolios. Notably, the model demonstrates robust downside protection during periods of heightened market stress, including the March 2020 downturn. These results suggest that reinforcement learning can enhance the responsiveness and stability of quantitative strategies under dynamic market conditions.
Open the resolved locator · 10.71465/fbf232
Metadata licence: OpenAlex metadata (CC0)