一种新的基于推荐的流媒体代理缓存替换机制  

A new proxy cache replacement mechanism of multimedia streams based on recommendation

作　　者：胡玉琦[1,2] 李晓娜[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004 [2]河北省计算机虚拟技术与系统集成重点实验室,河北秦皇岛066004

出　　处：《燕山大学学报》2015年第2期139-144,151,共7页Journal of Yanshan University

基　　金：教育部科技发展中心专项研究课题基金资助项目(2011109)

摘　　要：流媒体代理服务器缓存是有效缓解服务器负载、减少主干网传输的关键技术,而推荐算法是根据用户历史点播行为预测将被点播的可能性。现有的缓存算法没有考虑用户推荐对点播的影响,为此本文首先提出一种融合相异度的协同过滤推荐算法CFCD。其次,针对CFCD算法的推荐结果,定义反映流媒体文件被用户推荐程度的推荐度,并依此定义缓存价值函数,提出基于CFCD的流媒体代理缓存替换算法CRA_CFCD。仿真实验表明,CFCD算法能够提高流媒体对象的推荐精度,而提出的CRA_CFCD算法在缓存命中率和启动延迟方面改善了点播系统的性能。Streaming media proxy cache is the key technology of reducing load of streaming media server and transmission network. Recommendation algorithm is based on the historical access behavior of users to predict which media data would be probably re？quested in the further.The traditional cache replacement mechanisms have not concluded user′s recommendation influences.Firstly, CFCD （Collaborative Filtering Recommendation Algorithm Combined with Dissimilarity） algorithm is proposed.Secondly,Recom？mended Degree of a media file is defined based on the analysis of the CFCD recommendation result.And then CRA_CFCD（Cache Replacement Algorithm Based on CFCD） is proposed.Finally,Extensive simulations show that the CFCD algorithm can enhance the accuracy of the recommendations,CRA_CFCD is more effective in cache hit ratio and start-up latency.

关 键 词：代理缓存 缓存替换算法 协同过滤推荐算法 相异度 推荐度 

分 类 号：TP939.09[自动化与计算机技术]