Web-Site Boundary Detection Using Incremental RandomWalk Clustering

In this paper we describe a random walk clustering technique to address the Website Boundary Detection (WBD) problem. The technique is fully described and compared with alternative (breadth and depth first) approaches. The reported evaluation demonstrates that the random walk technique produces comparable or better results than those produced by these alternative techniques, while at the same time visiting fewer ‘noise’ pages. To demonstrate that the good results are not simply a consequence of a randomisation of the input data we also compare with a random ordering technique.