Applying an evolutionary algorithm to web search: The methodology of Evagent

An evolutionary algorithm is introduced to find authoritative resources on the Web. The problem of Web search is considered as an optimisation problem within hyperlinked space. We aim to find information that is both relevant andrecent so as to cope with the dynamic nature of the Web. Theoretical studies have been made on problem-specific search space, fitness functions and generic operators. The search space is constructed in the direction so the optimum driven by the reproduction operator with good hubs as a clue. Fitness functions combine text-based and link-based analysis. The (u+7) evolution strategy just implements the selection scheme of elitism. The mutation operator helps to prevent search from trapping in local optimisation by introducing multiple domains. Experiments have been performed to study algorithms performance. The algorithm has been implemented as a kernel component of an intelligent Web agent Evagent.