Agents, bookmarks and clicks
Mark Meiss, Bruno Gonçalves, José J. Ramasco, Alessandro Flammini, Filippo Menczer · 2010
Analysis has shown that the standard Markovian model of Web navigation is a poor predictor of actual Web traffic. Using empirical data, we characterize several properties of Web traffic that cannot be reproduced with Markovian models but can be explained by an agent-based model that adds several realistic browsing behaviors. First, agents maintain bookmark lists used as teleportation targets. Second, agents can retreat along visited links, a branching mechanism that can reproduce behavior such the back button and tabbed browsing. Finally, agents are sustained by visiting pages of topical interest, with adjacent pages being related. This modulates the production of new sessions, recreating heterogeneous session lengths. The resulting model reproduces individual behaviors from empirical data, reconciling the narrowly focused browsing patterns of individual users with the extreme heterogeneity of aggregate traffic measurements, and leading the way to more sophisticated, realistic, and effective ranking and crawling algorithms.