The rapid growth of technology has led to the creation of computing networks. The applications of the Internet of Things are becoming more and more visible with the expansion and development of sensors and the use of a series of equipment to connect to the Internet. Of course, the growth of any network will also provide some challenges. The main challenge of IoT like any other network is its security. In the field of security, there are issues such as attack detection, authentication, encryption and the so on. One of the most important attack is cyber-attacks that disrupt the network usage. One of the most important attacks on the IoT is BotNet attack. The most important challenges of this topic include very high computational complexity, lack of comparison with previous methods, lack of scalability, high execution time, lack of review of the proposed approach in terms of accuracy to detect and classify attacks and intrusions. Using intrusion detection systems for the IoT is an important step in identifying and detecting various attacks. Therefore, an algorithm that can solve these challenges has provided a near-optimal method. Using training-based models and algorithms such as Deep Dearning-Reinforcement Learning and XGBoost learning in combination (DRL-XGBoost) models can be an interesting approach to overcoming previous weaknesses. The data of this research is Bot-IoT-2018.