视频点播系统用户行为模型的构建与应用

Construction and Application of Users′ Behavior Model in the Video on Demand System

通过研究视频点播系统中用户的行为,构建出用户行为的数学模型,可以为视频媒体数据缓存算法改进提供依据.本文对常用的建立用户访问模型的数学模型进行了研究,对给出的一批媒体访问数据进行了建模分析,指出广延指数模型能够比Zipf模型更接近地描述实际影片点播的频率;并且用广延指数模型对影片播放长度进行了分析和拟合,验证了其有效性;进一步,基于所得的用户点播频率模型和点播长度模型,给出了一种缓存算法命中率上界的计算方法,从而为评价视频媒体数据缓存算法的性能提供了重要指导.

By studying the behavior of Video-On-Demand users,a mathematical behavior model can be bulit,which may guide the improvement of the video data caching algorithms.This paper investigates the common mathematical models of the users′ behavior,analyzes a batch of video data and finds that the stretched exponential model can better match the real demanding frequencies of videos than the Zipf model.It also implement the stretched exponential model to model the playback length of videos and gets very good fitting,which confirms the correctness of the obtained playback length model.Moreover,it proposes a method to predict the upper bound of the hitting rate of video data caching algorithms based on the achieved video demanding frequency model and the playback length model.That upper bound plays an important role in quantitatively evaluating video data caching algorithms.