Due to the recent prevalence of the Internet, some TV broadcasting services deliver videos using both electric wave broadcasting systems and the Internet (hybrid broadcasting environments). Video players encounter playback interruptions when they cannot receive a part of video data (video data segment) until the time to play it. The probability to encounter playback interruptions can be reduced by receiving video data segments earlier. However, it is difficult for video players to find from which reception channel (broadcasting system or the Internet) they can receive video data segments earlier since the time required for receiving them depends on various factors such as broadcasting schedules, the number of receiving video players, and so on. To find appropriate reception channels for reducing playback interruptions, we propose a system to select reception channel by machine learning.