Video Recommendations Based On User Behavior

The rapid development of 5G technology also brings the rapid development of information and Internet technology,which not only enriches people's life,but also changes people's life style.The increasing number of users who use video as entertainment increases the demand for video in the market.At the same time,in this era of "We-Media for everyone",more and more users are putting videos on the Internet.In the vast Internet resource information platform,personalized recommendation of video information has become an important means for the major platforms to retain customers.It is an indispensable and important technology in the field of video business to find the videos that meet users' needs and recommend to users in the resource database.This paper introduces in detail the commonly used recommendation technologies in the video recommendation field,among which collaborative filtering recommendation is a popular method in the video recommendation field,but there are some problems at the same time.Traditional collaborative filtering technology ignores the interaction information between users and projects,and lacks personalized recommendation for users.Therefore,it is of great significance to make personalized recommendation to users on how to make use of the characteristics of video and combine users' historical behavior information to find out the potential connection between them and solve the problem of missing interactive information in collaborative filtering.In view of the above problems and the sparse feature of user score matrix data in the recommendation system,a recommendation method based on video ontology isproposed in this paper.Firstly,the video ontology features are constructed and the similarity measurement of video ontology features is carried out.Secondly,the similarity of ontological features between unrated videos and videos with historical ratings is compared.Thirdly,multiple watched video scoring information with high similarity with the unseen video ontology was integrated to complete the rating estimation of missing data in the user-video scoring matrix,and the scoring matrix was filled.Finally,on the basis of model-based collaborative filtering recommendation,the improved matrix will select SVD model suitable for high-dimensional matrix,build the user rating recommendation model,and return the top-n movie recommendation results to the user.Through video platform climb took the actual data,verify the improved model in improving user ratings do have better effect on forecasting precision,based on the characteristics of video ontology recommended model is effective to a certain extent solved the user-video score data sparseness matrix data,and the user personalized requirements included in the model.