Internet video content and user experience has always been regarded as the core competitiveness of the video industry. Law of user behavior in the video content in-depth study of large-scale video service for the development of the Internet video industry is significant. In this paper, the Internet video is divided into 7 types of programs, and for the 9 kinds of user interaction behavior to do the correlation analysis. It is found that in the evening 8 o'clock to 10 PM primetime, TV series video programs have the largest number of sessions. In contrast, musical class video programs have the least number of sessions, but in each interactive behavior of the session, the number of interactive behavior is the most. For interactive behavior, the main user interaction behavior is fast forward and rewind operation. And in different content types, it is different. However other 8 kinds of interaction in a session may be in the same average number.