Design And Implementation Of Video Website Based On Personalized Recommendation

Due to the rapid development of network technology,the media form for people to obtain information through the network has changed from text and pictures to video.In addition,people also have a high demand for video-based social interaction,such as comment,real-time bullet screen,film review and so on.In addition,there are many kinds of videos on the Internet,so it is difficult for people to actively choose their favorite videos,so website developers need to recommend some videos that users may like to users.Using the rapid application development framework Spring Boot for back-end development,using the template engine Thymeleaf to render the website page,and using Hibernate to operate the persistence layer,the basic functions of the video website are realized,and the film review section is added to provide users with a series of interactive operations such as likes,comments,bulletins and film reviews.The bullet screen part adds a series of functions such as the search of the bullet screen list,the jump of the corresponding screen of the bullet screen,and the shielding setting of the bullet screen to optimize the user experience.Youtube Net recommendation algorithm is used for video recommendation,which is divided into two stages: recall and sorting.The input items of recall network and sorting network are modified and preprocessed according to the actual data of the website.In the recall stage,the recommendation problem is transformed into multi classification problem,and the video vector and user vector are obtained in the offline training stage,In the online stage,the inner product of the two is calculated as the video score,the videos are sorted according to the video score,and the first n videos are taken as the candidate set.In the sorting stage,on the basis of the recall stage,more features are added for training,and the predicted video viewing time of users is taken as the video score to sort the videos in the candidate set more accurately.The test shows that the website can run stably and meet the goal of design and implementation.Moreover,the videos recommended for users basically meet the user’s preferences,which can help users better find the videos they are interested in,and promote the published videos to the people who are most likely to be interested.