Abstract |
Little by little, from its birth up to now, the World Wide Web has progressively evolved to a large complex system with structure and properties that are formulated by the decentralized actions of tens of millions of individuals. Because of the uncoordinated nature of its growth, it was widely believed that the web lacks order and structure. Surprisingly, it was found that many properties of the web illustrate an intriguing regularity that describes its structure in a moderate way. Particularly, it was found that the web contains many small elements, but few large ones. A few sites are visited by millions of users per day, while most sites get a handful of visitors. In short, the distribution of the Internet users across Web sites is dramatically bumby. Albeit the fact that all web sites can be easily reached from everywhere and anybody, with no particular transportation or search cost, Internet users congregate into a few web sites, producing a strong agglomeration locational pattern. What may cause this concentration and how can we explain these sticking empirical data? In this thesis, we develop a percolation-like, agent-based computational model, towards the understanding of the formation of population agglomeration in the World Wide Web. We provide a theoretical framework and a methodology to explain this interesting system behaviour, based on the superposition of two interaction networks in which the decisions of individual agents are embedded in an environment of positive feedbacks and increasing returns. The model reproduces the concentration of population of the empirical data and demonstrates that a plausible explanation of Web agglomeration phenomena can lie on the assumptions of increasing returns and percolation networks with random connections.
|