Enhancing Financial Market Prediction with Reinforcement Learning and Ensemble Learning

Predicting financial market trends has been a complex and challenging task, even for experienced investors. Technical analysis is one of the commonly used methods by investors. However, machine learning models are now widely applied to predict stock prices and trends, among which reinforcement learning has received significant attention. Previous studies have integrated additional technical indicator features combined with historical price information to provide more information for reinforcement learning models, but the results have not been particularly outstanding. In this study, we adopt a special approach by categorizing technical indicators into two classifications: confirmation and prediction, each goes through a supervised learning model to generate BUY/SELL/NONE signals. We then experiment with ensemble learning methods to combine these two signal sources with the signal of the reinforcement learning model for the final prediction outcome. The experimental results show that integrating these two signal sources as input into the deep reinforcement learning model yields higher profits than the baseline model and achieves state-of-the-art performance in effectively integrating signals from technical indicators.