Research And Implementation Of Recommendation Algorithm For Video On Demand System

Online Video on Demand(VOD)service is a popular service in the Internet industry at present,and this service continues to expand rapidly with the development of the Internet.However,no matter what online video-on-demand platform,the number of videos displayed on the home page or navigation page is limited,which is in sharp contrast to the thou-sands of videos in the video library;at the same time,the VOD service often has a large number of users,and different users have different in-terest preferences.So how to show each user his favorite video in a lim-ited screen as much as possible has become an urgent problem to be solved.The emergence of the recommender system solves the above prob-lems well.Existing recommender systems generally follow the two-stage recall ranking process to generate recommendation results for users.The ranking problem is usually treated as a click-through rate prediction problem.However,the existing recommender system algorithms still have a lot of room for improvement in the field of VOD service,so this paper explores various recommender algorithms in the VOD scenario in detail and proposes improvements and innovations based on these exist-ing algorithms.This paper first explores the three collected data sets,including two public datasets in the field of recommender systems and one Internet Protocol Television(IPTV)dataset,and introduced the concepts of im-plicit feedback and explicit feedback in the recommender system.Next,this paper studies the existing multiple recall algorithms and improves the matrix decomposition algorithm among them.At the same time,this pa-per also proposes the graph query recall algorithm,and compares it with the traditional recall algorithm.Then,this paper explores the shortcom-ings of the existing ranking algorithm in the VOD scenario,proposes a user preference matching network(PMN)based on deep learning,and compares it with the existing mainstream ranking model on the three col-lected datasets,demonstrating the effectiveness of the PMN model pro-posed in this paper.