Research And Implementation Of A Portrait Short Video Recommendation System Based On Social Relationship

With the popularity of social applications and the rapid development of the short video industry,people are becoming more accustomed to sharing their lives and interested portrait or things with others by recording short videos on social platforms.People are also becoming more accustomed to watching short videos in fragmented free time to enjoy and relax.However,with the rapid increase in the number of users,it is bound to bring a problem of video overload,which will cause the users to spend a large part of the time searching for the video of their interest.Therefore,it is necessary to design and implement a short video recommendation system specifically for social platforms.Through such a system,autonomously recommended videos that may be of interest to user,so that the time for the users filtering the video is saved,which is more in line with the fast-paced life state of today's society,so as to provide a smarter and more humanized service.Based on the analysis of the short video features on the social platform,this paper finds that the portrait short videos are relatively large,and each video has a shorter duration and fewer characters appear in the video.Moreover,each user has a certain tendency to the characters in the video,so it is easier to obtain the video content and the user's preference orientation.As well as,by analyzing and summarizing the advantages and disadvantages of the existing video recommendation algorithm,combining the portraits that appear in the video with the user's social relationships,designs and implements a portrait short video recommendation system based on social relationship.First,by using face detection and comparison technology for video modeling to obtain the portraits and its' proportion which appearing in the video.As well as,through user behavior and social relationships for users modeling;Then merging the user's social relationship with the video content to obtain the user's nearest neighbor;Finally,recommend the video with the higher recommendation coefficient from the videos which are watched by the nearest neighbor recently.In order to construct a video recommendation system with application value,thispaper firstly has carried out a series of studies on the composition of the existing video recommendation system.Based on this,functional requirements analysis and non-functional requirements analysis were performed.For the above analysis,the front and back modules of the system are be designed and implemented.And with the improvement of the recommendation algorithm,it is more suitable for the recommendation of short video of social platform.Through comparison experiments,it can be seen that the accuracy of the proposed algorithm is improved compared with the existing mainstream algorithms.Therefore,this study has certain theoretical and practical significance.