The analysis of users' behavior of searching Webpages is the hotspot of current information searching. This paper focus on the weakness in the parallel calculation search of cloud calculation, like slow research speed, low efficiency and so on, a method based on Hadoop for mass users to search Web-pages is proposed, in which users' impact factors, time vector and Web-related factors are added to the algorithm based on the PageRank algorithm so as to further improve the efficiency for users in searching Web-pages. Analysis of query log in Youku laboratory is used in the experiment to prove algorithm in this paper has good effect as well as some guiding significance for users' behavior analysis in cloud computing.