Reinforcement Learning in Portfolio Management: A Comparative Study of DDPG and PPO

This research project aims to enhance portfolio management by incorporating modern investment strategies and cutting-edge technology. It focuses on optimizing portfolios through stock selection, building on the foundational work of Markowitz's mean-variance (MV) model. The project extends portfolio optimization to consider different asset time periods, risk measures, and high-quality asset selection. Additionally, it explores the online portfolio selection (OPS) model and its potential to maximize long-term wealth, incorporating concepts like universal portfolio (UP) strategies and transaction cost factors. Leveraging AI advancements, the project investigates the application of reinforcement learning algorithms, including deep Q-learning, proximal policy optimization (PPO), and deep deterministic policy gradient (DDPG), within portfolio management. This interdisciplinary endeavor aims to integrate finance, quantitative finance, and artificial intelligence to improve portfolio management methodologies and strategies.