Graphes du Web, Mesures d'importance à la PageRank

Abstract : The purpose of this thesis is to apply PageRank-like measures to Web graphs. The first part introduces the Web graphs. First we define the notion of indexable Web, then we give an insight on how big the effective crawls really are. Finally, we notice and use some of the structures that exist on the portions of the Web known as Web graphs. Then, the second part study deeply the PageRank algorithms. After a remainder on Markov chains theory is given an original classification of PageRank algorithms. From a basic model, we incorporate all the specificities needed to cope with real Web graphs. Lastly, new algorithms are proposed. BackRank uses an alternative random surfer modeling leading to a faster computation. The highly clustered structure of Web graphs allows a PageRank decomposition according to Web sites, and is the reason for introducing the algorithms FlowRank and BlowRank.
Document type :
Theses
Complete list of metadatas

Cited literature [61 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00667563
Contributor : Fabien Mathieu <>
Submitted on : Tuesday, February 7, 2012 - 10:42:11 PM
Last modification on : Friday, May 25, 2018 - 12:02:03 PM
Long-term archiving on : Tuesday, May 8, 2012 - 2:40:09 AM

Identifiers

  • HAL Id : tel-00667563, version 1

Citation

Fabien Mathieu. Graphes du Web, Mesures d'importance à la PageRank. Web. Université Montpellier II - Sciences et Techniques du Languedoc, 2004. Français. ⟨tel-00667563⟩

Share

Metrics

Record views

1387

Files downloads

5095