Abstarct: Stock, currency pair, and cryptocurrency trading have consistently attracted both experts and non-experts due to their high appeal and profit potential. In 2023, the daily trading volume in the world's largest financial markets, such as Forex and Nasdaq, reached $6.6 trillion and $300 billion, respectively. This staggering volume has led to a growing number of traders. One of the main challenges for traders is finding profitable strategies in volatile markets. Recent advancements in artificial intelligence, particularly in machine learning and reinforcement learning, have made it possible to develop intelligent trading strategies. These advancements, especially in deep reinforcement learning, allow the use of algorithms like A2C, PPO, DDPG, and TD3 as intelligent agents. These agents, by interacting with and adapting to turbulent market environments, help improve trading outcomes and maximize profits. In this research, we implemented deep reinforcement learning algorithms to assess the advantages and limitations of each agent. Then, by introducing a two-laxyer ensemble strategy baxsed on voting among all agents, we achieved maximized profits and reduced risks in financial trading. This method was applied to 30 leading stocks from Dow Jones, Nasdaq, and S&P 500 with a one-day trading window. The results demonstrate that the proposed strategy achieved a 51% return in Dow Jones, 59% in Nasdaq, and 49% in S&P500, outperforming machine-learning-only methods.