Research On The Technologies On The Video Recommendation Based On Social Network

Currently, Weibo is one of the most popular social networking websites, video watch as one of its most important user experience, users can watch the video which recommended by their friends through Weibo social circle. In order to enable users to watch the video they are interested in, video recommendation system becomes a part of the social network. Nowadays, the researches on video recommendation system mainly depending on the amount of video and video rank, as well as the video which released by their friend, it can enhance the experience of the users to watch video in certain degree. Therefore, how to more effectively recommend the interesting video for users, which becomes a popular research issue on video recommendation system.This paper mainly focus on the problem that the video which recommended by Weibo system is unlikely to meet the needs of users. So, we study the video recommended method from user recommendation model and trust evaluation model respectively, and proposed the user recommendation model based on trust and the video trust evaluation model based on content.The user recommendation model based on trust considered three aspects,including user similarity, friend relationship and user interaction, to find more potential friends for the target users. Calculation of user similarity was based on semantic similarity tag. Friend relationship considered attention, mutual powder,celebrity and fans. And interaction considered three interaction factors, including forwarding, praise and favorites level. The model of video quality evaluation considered video ratings and video activity level. Among them, the video ratings were based on the information between users and video. The video activity level referenced evaluation method from user activity level, evaluated from three aspects including video forwarding level, praise and favorites level. Users can find out the users similar to target users through user recommendation model and can guarantee the quality ofthe video through video evaluation model. So, users can receive more user interest coincide with the target video.Finally, we conducted experiments compared with TBR-d and Mwalker respectively. We adopted three evaluation functions, including accuracy, recall and F1 function, to compare the three algorithms performance, and results showed that the algorithm we proposed performance better than other two video recommendation algorithms in the effect of the recommendation. So, the feasibility and effectiveness of our proposed algorithm was verified.