摘要
传统的Web预取机制所采用的预测算法主要针对用户个人预取,随着用户数量激增会加重网络负载,降低网络资源利用率。针对这一问题,在综合分析经典ART1神经网络模型和预取系统的基础上,提出一种基于用户聚类的UCPM模型预取新方法。首先,基于改进的ART1算法对用户访问序列特征向量进行聚类,挖掘兴趣相似的用户集合;然后,针对自底向上权重最大值所关联用户群的兴趣进行预取;最后,在Web预取系统上验证该方法的有效性及可靠性。实验结果表明,UCPM模型表现出较好的聚类效果,同时应用在预取系统保持了较高的预测准确率,降低了延迟比和流量开销比。
Prediction algorithm of traditional prefetching mechanism adopted by the invisible network load will increase the utilization rate,reduce the cyber source rate. In view of this,through the comprehensive analysis of the classical ART1 neural network model and the prefetching system,a new method of UCMP prefetching model is proposed based on user clustering. Firstly,based on improved traditional ART1 algorithm for clustering user access sequence feature vector,mine the similar interests of a user set,and then prefetching technique predicts the destination pages for user set on the basis of bottom-up maximum weight value. Finally,the experiments on Web prefetching system verify the effectiveness and reliability of the proposed algorithm. Experimental results show that the model exhibits better cluste-ring effect,at the same time,the prediction accuracy in prefetching system has been retained highly,lowering the ratio of latency and traf-fic.
出处
《计算机技术与发展》
2015年第9期106-110,共5页
Computer Technology and Development
基金
国家自然科学基金资助项目(U1304603)
2012年郑州市科技计划项目(121PPTGG364)
