Integrating Weak Aggregating Algorithm and Reinforcement Learning for Online Portfolio Selection: The WARL Strategy

Online portfolio selection is an open and fundamental problem that has been attracting the attention of numerous researchers. Our aim is to address the challenges posed by the dynamic and complex financial market environment, where single investment strategies often face difficulties in adapting to rapid changes and unexpected events. Thus, we introduced the WARL strategy for online portfolio selection, which aggregated a heterogeneous expert set using a meta-learning approach called the weak aggregating algorithm (WAA). We replaced the online decision-makers in WAA with a reinforcement learning (RL) agent and proposed a novel reward function accounting for multiple objectives and event-triggering mechanisms. Based on real-world data numerical experiments, the WARL strategy outperforms the comparison strategies in a multi-metric evaluation. The comprehensive analysis conducted in this study supports the conclusion that the WARL strategy serves as a dependable and adaptive investment option, offering investors an effective means to optimize portfolio returns while effectively managing risks.