A Hybrid Semantic Matrix Based Video Recommendation System

With the rapid development of the Internet, the content of web sites increased dramatically. How to get the interested information quickly has becoming an important issue. Especially in the video service field, the users don't know exactly the name of the video in which they are interested, or they don't have a strong willingness to browsing video site. How to recommend the potential interested videos to users is becoming more and more important for the video service providers. Therefore, the video recommendation system becomes a hot topic both in academy and industry. For a personalized video recommendation system, it's really important to recommend interested videos effectively and efficiently to users according to their special interests and browsing history, which is the focus of this study.Based on the backgrounds mentioned above, existing technologies in research field and the actual requirements of the personalized video recommendation system,main works completed in this paper are listed as follows.First, to address the existing drawbacks in content-based recommendation algorithms and collaborative filtering recommendation, this paper proposes an new video recommendation algorithm with intelligent feature extraction, label quantification and the method of weight distribution. The proposed algorithm can solve partly the problem of over-reliance on expert experience, and improve the accuracy of the recommendation system.Then, through improving the original video content-based recommendation algorithm, this paper proposes a video recommended method which is based on a hybrid semantic matrix. Moreover, we design and implement a video recommendation system which is based on the proposed recommendation method.Finally, it presents the system in detail, from requirements analysis, architecture design, to the full realization of process. The experimental results show that an improved precision and recall measurement can be obtained.