Weighted PageRank: Cluster-Related Weights

Abstract

PageRank is a way to rank Web pages taking into account hyper-link structure of the Web. PageRank provides efficient and simple method to find out ranking of Web pages exploiting hyper-link structure of the Web. However, it produces just an approximation of the ranking since the random surfer model uses just uniform distributions for all situation of choice happening during the surf process. In particular, this implies that the random surfer has no preferences. The assumption is limited by its nature. Personalized PageRank was designed to solve the problem but it is still quite restrictive since it assumes non-uniform preferences just at jumping to arbitrary page on the Web and non-preferring behaviour when following outgoing hyper-links. Taking into account these limitations and restrictions of PageRank and Personalized PageRank we propose Weighted PageRank where we are free to weight hyper-links according any possible preferring behaviour of a user. In particular, cluster-related weights are considered.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 01, 2008
Accession Number
ADA512719

Entities

People

  • Danil Nemirovsky
  • Konstantin Avrachenkov

Organizations

  • Saint Petersburg State University

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Abstracts
  • Algorithms
  • Clustering
  • Information Operations
  • Instructions
  • Markov Chains
  • Probability
  • Probability Distributions
  • Standards
  • Stationary
  • Transitions
  • Universities

Fields of Study

  • Computer science

Readers

  • Geospatial Intelligence and Artificial Intelligence Analytics
  • Graph Algorithms and Convex Optimization.
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.