The inherent complexity and ever-changing nature of financial markets make it challenging to make investment decisions. Allocating investments among different assets is a popular strategy for minimizing risk and potentially increasing gains, but finding the optimal combination of assets at any given time is a difficult task.

Recently, trading decision systems based on Reinforcement Learning (RL) methods have been applied in portfolio management and are gaining popularity.

In this work, we explore the effectiveness of three model-free RL algorithms for navigating financial markets and finding optimal trading strategies, while accounting for historical prices, volatility, and transaction costs. To achieve this, we would use trading agents based on DQN, PPO, and A2C algorithms and evaluate their performance on historical data from the USA, European and Asian financial markets with quantitative trading metrics such as final Accumulated Portfolio Value, Sharpe ratio, Maximum Drawdown, etc., and provide a conclusive report on the effectiveness of these algorithms for automated portfolio allocation.