摘要
提出一种面向园区网VOD代理系统的节目分布策略,首先构建基于BP神经网络的弱分类器并采用AdaBoost算法将其提升为强分类器,然后基于该分类器对系统中的节目按流行度高低进行分类,再根据分类结果采用基于分组和点播率的方法对系统中的代理服务器进行文件分布.该策略充分利用用户点播行为中蕴含的信息,并具有执行结构简洁,易于实施的优点,仿真实验结果表明该策略能有效提高代理服务器的利用率.
This paper presents a data distribution strategy for VOD proxy system in campus networks. First, the weak classifiers were constructed based on BP neural network, then the weak classifiers were ＇combined and upgraded to strong classifier by using AdaBoost algorithm. The strong classifier was used to classify the programs in VOD system according to the prevalence. The classification results were used to distribute the files in the proxy servers based on program group and rate of request program. The strategy makes full use of the information inherent in the behaviors of users in request program. It has simple structure and can be easy to be implemented. The simulation results show that strategy can effectively improve the utilization of the VOD proxy servers.
出处
《浙江工业大学学报》
CAS
北大核心
2010年第2期182-185,共4页
Journal of Zhejiang University of Technology
基金
番禺区科技攻关项目(2008-Z-54-1)
广州市高校2006年科技计划资助项目(62077)
