This paper focuses on enhanced channel planning for WI-FI systems. Based on a real measurement of the frequency spectrum background, the quality of each channel has been classified. Machine learning algorithms are used to process these data and control the system. Reinforcement learning (a special example of machine learning) is used due to its complexity as a test-trial agents system. In the proposed system punishment and reward schema are utilized making it possible to change channels during the transmission and use the one with minimal interference. The promising increase of SINR up to 5% for outdoor scenarios and up to 30% for indoor scenarios should be applied to improve modulation schemas and increase data throughput.