Abstract:Goal-based investing is an approach to wealth management that prioritizes achieving specific financial goals. It is naturally formulated as a sequential decision-making problem as it requires choosing the appropriate investment until a goal is achieved. Consequently, reinforcement learning, a machine learning technique appropriate for sequential decision-making, offers a promising path for optimizing these investment strategies. In this paper, a novel approach for robust goal-based wealth management based on deep reinforcement learning is proposed. The experimental results indicate its superiority over several goal-based wealth management benchmarks on both simulated and historical market data.
What problem does this paper attempt to address?
The paper primarily focuses on addressing the investment strategy optimization problem in Goal-Based Wealth Management (GBWM). Specifically, the researchers propose a new method based on Deep Reinforcement Learning (DRL) aimed at optimizing goal-oriented investment strategies through machine learning techniques.
Traditional wealth management typically emphasizes balancing expected returns and risks, whereas GBWM is more concerned with maximizing the probability of achieving specific financial goals. For example, saving to pay for tuition, retirement, or purchasing a property. As the target date approaches, asset allocation gradually becomes more conservative to reduce risk. However, traditional fixed glide paths may not be the optimal choice, especially when the target date is near but the goal has not yet been achieved, requiring more aggressive investment strategies.
To address the above issues, the researchers developed a DRL-based framework that can adaptively adjust investment strategies to increase the likelihood of achieving the goals. They utilized the Proximal Policy Optimization (PPO) algorithm and designed a feedforward neural network with two hidden layers to approximate the policy function. Additionally, the researchers employed various training and testing procedures to ensure the robustness and generalization capability of the proposed method, including historical data testing, simulated data testing, and bootstrap data testing.
Experimental results indicate that the proposed DRL method outperforms several commonly used GBWM benchmark methods on both simulated market data and historical market data. These benchmark methods include deterministic glide paths, Merton’s Constant, Variance Budgeting, and dynamic programming.
In summary, the goal of this research is to improve the investment decision-making process in GBWM through DRL technology to increase the probability of achieving specific financial goals. The experiments demonstrate the effectiveness and advancement of this method.