When we broadcast web pages package through broadband broadcast network, what we will broadcast and how we broadcast are two problems. First, we formalize these problems as a clustering problem. Second, we use the association rule to give a better result. In this paper we provide a new web mining method WebClustering to resolve these problems. This method combines the clustering idea and association rule discovery idea. The mining object is the web pages of Cache and Log in WWW Proxy server. By this method we can find a valuable Web broadcast set and create some index HTML pages to direct the users to navigate.