摘要： 针对流行度随时间变化的特性，利用回归分析技术给出了一种流媒体文件的流行度预测算法，并在增加少量存储空间及计算时间消耗的情况下，将该预测算法应用于流媒体代理缓存服务器的缓存替换算法之中，模拟实验表明，该方法能减少缓存的替换次数，提高缓存命中率，性能较优。

关键词: 流媒体, 流行度预测, 代理缓存, 缓存替换算法

Abstract: Caching frequently used streaming media files at proxies close to client is an effective means to reduce the client startup delay, the server traffic and the network bandwidth consuming. Using regression models, this paper puts forward a new prediction method of popularity prediction of streaming media file, based on this method, a cache replacement algorithm of the proxy is given. The simulation experiments indicate that this algorithm can make the caching system maintain better performance.

Key words: Streaming media, Popularity prediction, Proxy cache, Cache replacement algorithm