基于矩阵聚类的页面预测研究  

Research of Web Page Predicting Based on Matrix Cluster

作　　者：彭玉青[1] 田海山[1] 陈美娜[1] 

机构地区：[1]河北工业大学计算机科学与软件学院,天津300130

出　　处：《计算机工程》2004年第8期156-158,共3页Computer Engineering

摘　　要：提出一种基于Web日志的分析方法，构建UserID-URL关联矩阵，通过矩阵聚类进行页面预测，为了加快聚类速度和效果，引进Ping-Pong算法，实验表明基于矩阵聚类的页面预测达到较高的命中率。In this paper, a new method of analysis on Web log is proposed. A UserID-URL matrix is created and it achieves Web page predicting through matrix cluster, additionally a Ping-Pong algorithm is introduced in order to increase the speed of cluster. The experiments show that the Web page predicting through matrix cluster has high hit probability.

关 键 词：矩阵聚类 Ping—Pong算法 页面预测 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]