CASoRT系统中基于聚集特性的在线流行度预测方法

少数在线热门内容会在短时间内吸引大量用户的访问，并占用大量的网络传输资源。如果能预知内容的热门程度（即流行度）并将热门内容广播给潜在用户，将极大地节省网络传输资源，这正是CASoRT 系统的主要功能。通过对国内商业蜂窝通信系统中收集的相关数据进行分析和研究，发现在用户行为、地理位置、数据内容等方面存在明显的聚集特性。根据上述特性给出了两个流行度预测算法，即对数线性和恒定比例模型，并使用最优观察门限改善两算法的性能。通过对两算法仿真结果的比较，对数线性模型表现更优，被选作系统的在线流行度预测方法。

Online Popularity Prediction Scheme Based on Converging Property in Content Aware Soft Real Time Media Broadcast System

A small number of online popular contents are often clicked by a great quantity of users in a short period,and take the most of the wireless cellular network traffic.With popularity prediction,the popular contents would be broadcasted to the potential users for saving a lot of transmitting resource,as illustrated in content aware soft real time media broadcast（CASoRT）system.With the data set collected from the Chinese commercial cellular network,the converging property of web contents,users and geographic positions in online news was shown.Then,two prediction schemes such as linear log and constant scaling model were proposed to estimate the popularity of online news,and improved by an optimal observation threshold.After comparison of simulation results,the linear log model performs better.