Lighting control systems have been developing and improving in the last 10-20 years. The smart lighting market will only grow shortly. Traditional control algorithms (schedule, astronomical schedule, motion and presence sensors) cannot provide either the required levels of visual comfort or the optimal energy consumption of the lighting system. In addition, it should be noted that there is a tendency for modern users to increase the requirements for personalizing lighting conditions and flexibility in changing them. There are several directions for developing lighting control systems: biodynamic lighting, reducing energy consumption, focusing on visual comfort. The article proposes to consider machine learning as a promising and new approach in lighting control systems. The existing lighting control systems and scientific research in this area are analyzed in terms of their orientation. Machine learning has several advantages and disadvantages. Its initial implementation is quite complicated, but it is possible to simplify the setup and operation of lighting control systems significantly with its further use.