DRL Optimizes Investment Portfolios for Return and Downside Risk.

Portfolio optimization in uncertain markets is a complex multi-objective problem, balancing return, risk, and various constraints. Existing static optimization methods often fall short in capturing sequential decision-making, tail risk, and market frictions like transaction costs. To address these limitations, researchers developed MORP-DRL (Multi-Objective Reliability Based Deep Reinforcement Learning), a novel framework for portfolio optimization. MORP-DRL employs a deep reinforcement learning approach, specifically using Proximal Policy Optimization (PPO), to simultaneously optimize expected return and minimize downside risk. It incorporates three complementary risk measures: variance, Conditional Value-at-Risk (CVaR), and Entropic Value-at-Risk (EVaR). To model market uncertainty and heavy-tailed behavior, asset returns are simulated using GARCH(1,1), Extreme Value Theory, and t-copula dependence structures. Experiments across different market regimes (pre-COVID, COVID, post-COVID) on ten global equity indices demonstrate that MORP-DRL achieves competitive risk-return performance, significantly reduces downside risk during volatile periods, and scales effectively to high-dimensional portfolios, outperforming traditional methods like NSGA-II.