Markov Chain Methods in Ranking Webpages and Websites

Zhi-Ming Ma
Institute of Applied Mathematics, Academy of Math and Systems Science, Beijing 100080, P. R. China


Abstract

The talk is based on our recent joint work with Microsoft Research Asia. We adopt the random walk to describe the behavior of Web surfers and show that PageRank evaluates the importance of webpages by the mean frequency of the Markov chain visiting webpages. Following the same idea, we propose AggregateRank to evaluate the importance of websites. We construct a returntime Markov chain to describe the probabilistic relation between the importance of webpages and that of websites. We propose also the AggregateRank algorithm and discuss the convergence speed and the corresponding error bound.