Video Multicast Caching And Rating Prediction

In recent years,global mobile data traffic increases at an explosive speed.Video data traffic occupies a lot of it.Requests for videos from people during peak hours cause network traffic jam.For the purpose of alleviating network traffic load during peak hours,reducing bandwidth that video transmission occupies has been a critical problem.As the price of storage becomes lower in recent years,we can cache a fraction of video file or the whole file in the local storage in advance.As a result people can get their request video file from local storage.This method is very promising for alleviating network traffic jam.If combined with multicast the server only needs to occupy one piece of bandwidth resource and sends the video file to multiple users.This will lighten pressure of video transmission for the network greatly.If users want to get their request video files during peak hours from local storage,it will be necessary to predict the requests of users in advance and cache the video files users likely enjoy.Taking use of recommendation algorithms to predict the requests of users is an important basis of cache working.In this paper,we take decentralized coded caching algorithm as our theory basis.And we use USRP(Universal Software Radio Peripheral)as hardware platform and GNU Radio as software tool to set up a video caching and multicasting system.We take advantage of the system to explore the benefit of the algorithm can bring to the wireless video transmission.Compared to previous systems,the system adopts our self-designed data structure for the purpose of the system working in physical layer.The experimental result shows that decentralized coded caching can lessen the pressure of video transmission.The performance of the system is close to the theoretical result.Next we predict the ratings of totally unrated movies.We use textual analysis of the movie plots instead of rating matrix from users to extract movie features.Next we use collaborative filtering to predict ratings of the movies.The experimental results show that compared to other algorithms the algorithm can improve the accuracy of prediction to a certain extent.