Traditional strategies for managing portfolios often rely on rigid models and people’s gut feelings, which might not work as well when the market is changing and unclear. Instead, DRL seems like a good way to make decisions that are adaptable because it learns from past data and interacts with the market environment. A system that combines DRL algorithms with financial data is what we’re suggesting as the best way to make trade and stock allocation decisions. By changing stock weights based on market signs and past success, the DRL robot learns how to get the most total benefits. We show that our method works by doing a lot of back testing on old market data and comparing it to other strategies that have been used before. In terms of risk-adjusted returns and ability to handle market instability, our DRL-based portfolio management plan does better than standard approaches. We also look at how different hyper-parameters and design choices affect how well the DRL agent works. Overall, our study adds to the growing amount of research on using machine learning in financial markets and gives practitioners who want to improve portfolio management through automation and flexible decision-making useful information.