Research On Clustering Incorporation Based Recommendation Techniques For Mobile Streaming Recommendation And Caching

Video streaming has always been the killer application on the Internet,and has been occupying a large number of network traffic,the optimization of video streaming service is a hot topic in the academic world.Recent years as the popularity of mobile equipments and the progress in wireless transmission,users now get more freedom in streaming service:On one hand,they can now request streaming service anywhere in anytime,the demand for video service increase;On the other hand,users have the abil-ity to upload videos,which makes the number of videos on the web increase rapidly,the need for personalized service also appear accordingly.So,personalized preference of users need fully exploiting,to enhance the quality of personalized service;At the same time,the exploition of users' preference offer a new chance for enhancing quality of service and video transmission.To solve the above problems,we suggest enhancing the bayesian ranking method by utlizing the clustering signal in video viewing session,and get the multiple-signal combined recommendation model.With the recommendation model,we build a caching mechanism based upon" smartphones oriented collaborative mobile streaming media service system".By ensuring the cache hit ratio,the mechanism uses broadcast and overhearing to saving the transmission load in Ad-hoc network.Our works include following 2 parts:1.multiple-signals combined recommendation model.In video recommendation application,besides the users' preference information,there are also clustering signal,namely,users tend to watch the same kind of videos in the same video watching session.We try to utilize these information to enhance traditional pref-erence based method,and propose the multiple-signals combined recommenda-tion model.To verify the effectiveness of our method,we run experiment on real dataset,and demonstrate that clustering information can effectively improve the performance of video recomme-ndation.2.clustering recommendation based video caching mechanism.To cache short videos,the first step will be inducing users' preference and predicting which videos users are likely to watch(mobile phones have enough storage to cache some short videos now).To do this,we use recommendation model to guess users' preference,and cache videos with high preference score.This gurantee us with a enough cache hit ratio.At the same time,we build a recommendation based caching model on our video streaming system.By utilizing the broadcast characteristic of wireless transmission,the device can overhear the data block,hence to achieve intension that "one transmission satisfy many requests".Con-sequently,it alleviate the channel collision problem in Ad-hoc netwrok.