ProfitPulse: Reinforcement Learning-Driven Trading Strategy

This paper presents a robust investment strategy using reinforcement learning to maximize total wealth. The proposed strategy is evaluated against traditional methods, including "buy-and-hold" and moving average convergence divergence, with a comprehensive framework addressing key factors in price determination. Optimization in decision-making of reinforcement learning aims at maximizing the cumulative rewards that could be attained in uncertain environments. The approach focuses on the development of a simulated trading environment with realistic factors such as price movements, transactions, and capital constraints. Through historical data and price movements, trading volume, and financial news, a model is trained through Q-Learning and Deep Q-Networks to infer optimal trading policies. The system features a user-friendly frontend developed using Streamlit for interaction and visualization. This study demonstrates the effectiveness of reinforcement learning in developing adaptive, data-driven trading strategies, which are shown to perform better than traditional methods for financial forecasting and decision-making.