Abstarct: The forex market is the largest international market for currency exchange, and individuals from around the world use this market for currency investment. Since the formation of this global market until today, providing a profitable solution for investing in this market has been one of the most significant challenges for investors and analysts in the capital market domain, aiming for higher profits in this market. In recent years, machine learning algorithms, especially deep learning, have effectively performed operations such as data categorization or feature extraction baxsed on intelligent learning from previous data in various fields. Additionally, with the introduction of reinforcement learning and intelligent agents capable of taking actions in different environments without the need for previous data, using algorithms such as Q-Learning has created a different approach to learning and acting in diverse environments. In this study, an attempt has been made to use an intelligent agent trained on deep learning and Q-Learning algorithm in reinforcement learning to conduct transactions in the forex market, aiming to achieve the highest profit and the least loss during the testing period. This intelligent reinforcement agent, with a trend-baxsed trading approach using various machine learning classification methods in the forex market for multiple distinct currencies in a one-day time frxame, is compared. It is evident that these proposed intelligent agents outperform previous structures, achieving higher profitability percentages of 64.9% and 38.5% for the aud/usd and eur/usd currencies, respectively. Furthermore, in general, intelligent agents demonstrate higher profitability compared to machine learning-baxsed trading approaches.