Incremental iteration method for fast pagerank computation

Abstract

As the search services have been widely available on the web, page-ranking algorithm gained great attention in the recent decade. PageRank is the most popular ranking scheme that is currently well-known even to the public. PageRank uses a hyperlink matrix which represents the whole web structure and the size of the web is incredibly large in general so that a fast calculation method is needed to efficiently compute an enormous number of page ranks on the web. In this paper, we propose a new PageRank computation method, incremental iteration method in order to considerably reduce total computational cost. Our method makes good use of faster convergence procedure of power iteration than conventional one. Additionally, our method can be effectively combined with other conventional methods for more reduction of computational cost. In experiment, we demonstrate efficiency and effectiveness of our proposed method.