An Investigation of Current Applications in Quantitative Finance Based on Reinforcement Learning and Supervised Learning Methods

The rapid evolution of quantitative finance and the growing complexity of modern financial markets have revealed the limitations of traditional statistical and rule-based trading systems. Reinforcement Learning (RL) and Supervised Learning (SL) have emerged as promising alternatives, offering data-driven approaches to prediction, optimization, and strategy design. Unlike SL, which focuses primarily on forecasting returns through labeled data, RL leverages trial-and-error interactions with dynamic environments to optimize decision-making, making it especially suitable for tasks such as order execution, portfolio allocation, and risk control. This paper provides an exhaustive and accessible overview of RL and supervised learning in various financial fields. It examines methodologies ranging from meta-reinforcement learning frameworks and multi-agent trading systems to supervised models enhanced by technical indicators, sentiment recognition, and anomaly detection. While these approaches demonstrate improved profitability and adaptability, they face challenges including high computational complexity, sensitivity to hyperparameters, reliance on historical data, and limited interpretability. Furthermore, generalizability across non-stationary market conditions remains an open issue. Looking ahead, future directions involve integrating alternative data sources, adopting hybrid RL-SL frameworks, enhancing explainability, and extending applications to diverse asset classes. By highlighting both opportunities and limitations, this review outlines a research roadmap for advancing intelligent and robust quantitative trading strategies.