Fast Mining And Prediction Of Hot Video In Video Sharing Website

With the development of the Internet,online-video has become one of the most important industries in the internet.At present,online-video have huge size of the number of users,their morphological diversity and the amount of data characteristics make their influence has gone beyond the traditional media such as television and print.With the Web2.0 development opportunity,the Internet is becoming more open and more wild,the video sharing site to individual users for network video release source,the program content to maximize the rich,the number of videos also showed explosive growth.Video sharing sites like Youku,youtube and other online-video site added social elements,making the video sharing website can reflect more content elements,and also to provide a rich metadata for hot-spot,the network of public opinion and the research direction.Popular video can bring huge traffic to the site,to attract more users to watch,and also can lead users to comment on the video content,to express their views.Therefore,how to share the popular video resources potential rapid discovery site in the video,and video tracking and monitoring heat changes,is a problem to be solved urgently.The focus of this paper is based on the video sharing site characteristics,combined with gray Verhurlst prediction model,establish a popular video mining prediction model,to explore the development trend of popular video quickly and prediction how it goes.Firstly,according to the video sharing website fans social characteristics,analysis the video sharing site on the video popularity several factors,and puts forward the quality to account,level the number of fans,the topic of heat are factors that can affect video popularity,in the popular video mining,we account quality and topic of heat of the two factors as parameters added to the calculation of heat,compared with the traditional hot video mining method,shorten the popular video to explore the time,to improve the efficiency of the popular video mining.Then in this paper,through the comparison of the existing trend prediction model of the advantages and disadvantages,we chose grey Verhulst model as the basic model,then add our hot topic parameters,compared two models of video popularity prediction,eventually through experiments proved our improved model improves the prediction accuracy,the error rate reduced.Finally,we design a video website hot spot discovery and prediction system based on the theoretical model,the system's performance is well.Proved the feasibility,effectiveness and reliability of our model.