基于有向带权图的页面聚类算法研究

Study on Page Clustering Algorithms Based on Weighted Directed Graph

聚类算法是数据挖掘中的一个重要的分析工具。Web使用挖掘中的聚类分析一般分为用户聚类和页面聚类。其中页面聚类是指导网站结构离线优化的重要方法。利用有向带权图表示用户的访问会话记录,对建立的有向带权图模型运用聚类算法实现页面聚类。选取真实数据对典型的聚类算法K-means算法、DBSCAN算法和COBWEB算法进行实验。实验结果表明,在选取的数据集范围内,COBWEB算法准确率要高于K-means算法和DBSCAN算法,时间性能与用户访问频率矩阵大小有密切关系。

Clustering algorithm is an important analytical tool in data mining. Clustering analysis is generally fallen into user clustering and page clustering in Web usage mining. Page clustering is an important methods for guiding for the structure of the site off- line optimization. This paper use weighted directed graph to describe user visit and conversation records, and use clustering algorithms to realize the page clustering by the weighted directed graph mode established. Select the real data carries on the experiment to the typical clustering algorithms K-means algorithm,DBSCAN algorithm and COBWEB algorithm. The experiments results indicate that in the selected data sets, the accuracy rate of COBWEB algorithm is higher than that of K - means algorithm and DBSCAN algorithm, and the time capability is closely related to the size of user visit frequency matrix.