Optimal Prefetching in Random Trees

We propose and analyze a model for optimizing the prefetching of documents, in situation where connection between documents is discovered progressively. A random surfer moves along edges tree representing possible sequences which known to controller only up depth d. quantity k can be prefetched two movements. The question determine nodes should so as minimize probability moving node not prefetched. analyzed with tools Markov decision process theory. formally identified optimal policy several situations, we it numerically others.