Research On Caching Schemes Based On Content Prediction In Ultra-Dense Network

The dense deployment of small cells in ultra-dense network is an effective way to meet the surging demand for mobile data.But the data rate is severely restricted by the limited capacity of backhaul link at peak time of traffic.Storing popular files on small cells can alleviate the pressure on the backhaul link and reduce the content acquisition delay,with serving local users directly.Proactive cache scheme is a way to cache popular content in advance of peak time of traffic,which not only alleviates the pressure on the backhaul link at peak time of traffic,but also makes full use of the bandwidth resources at off-peak time of traffic.Due to the time-varying popularity of content(e.g.,video files)in current network environment,how to predict the popularity of content accurately becomes a key to improve the performance of the proactive cache scheme.For this problem,this paper studies the caching schemes from the perspective of user mobility.Firstly,a cache decision scheme based on user request prediction is proposed for small cell scenario with low mobility of users.At first,according to the user's historical request content,the scheme adopts the k-mediods clustering algorithm to classify the users under overlapping coverage of small cells,and the same kind of users will access to the same small cell.And then,the collaborative filtering algorithm is used to predict the user's future request,and the content popularity is calculated so as to make cache decision of the small cell.Compared with the existing Least Frequently Used(LFU)strategy and random strategy,simulation results show that the proposed scheme improves the cache hit ratio and reduces the content acquisition delay.Secondly,a cache replacement scheme based on content popularity prediction is proposed for small cell scenario with high mobility of users.At first,the number of content request is predicted by using the Auto Regressive Integrated Moving Average Model.And then,considering the number of request and the interval of the last request for the content,the content popularity prediction model is established.When the cache space of small cell is insufficient,the content with low prediction popularity will be replaced by the content with high prediction popularity,with increasing cache space utilization rate.Compared with the existing Least Recently Used(LRU)strategy and LFU strategy,simulation results show that the proposed scheme improves cache hit rate and reduces the content acquisition delay as well as core network traffic.