摘要
现有视频点播系统的用户行为建模研究仅从会话的角度考察视频交互式请求数分布模型.提出从视频对象的角度考察用户交互式请求数分布.观察到交互式请求数的分布表现出重尾现象;通过对实际用户访问数据的统计分析,证明常用幂律模型不适合刻画交互式请求数分布;提出采用广延指数模型对其建模.对不同时间区间内用户访问数据的分析表明,广延指数模型较好的描述了交互式请求数的分布,其形状参数描述了用户交互式请求的波动性,尺度参数则刻画了用户交互式请求的时效特性.
Current study on modeling of user behavior in video on demand system only investigate the distribution of number of user interactive requests per session.The modeling and analysis of distribution of number of interactive requests per video were proposed.The distribution of number of interactive requests has been observed to appear to heavy tailed.Statistical analysis on a real workload showed that the common power law model may not be suitable to characterize the distribution.The stretched exponential model was proposed.Numerical studies on the real workload in different time scales indicated that the stretched exponential model well characterized the distribution of number of user interactive requests.The shape parameter of the stretched exponential model characterizes the volatility of user interactive requests and the scale parameter characterizes the aging effect of user interactive request.
出处
《小型微型计算机系统》
CSCD
北大核心
2010年第7期1426-1432,共7页
Journal of Chinese Computer Systems
基金
国家科技支撑计划(2008BAH28B04)资助
安徽省高校省级自然科学基金项目(KJ2008A106)资助
关键词
交互式请求
用户行为模型
广延指数模型
重尾
时效特性
interactive request
user behavior model
stretched exponential model
heavy tail
aging effect
