TNA010 - PageRank Algorithm

Overview

This project implements the PageRank algorithm, originally developed by Google, to rank web pages based on their importance in a network. The implementation is based on the methods described in Matrix Methods in Data Mining and Pattern Recognition (2nd Edition) by Lars Eldén.

Algorithm Description

PageRank works by modeling web surfing behavior as a probabilistic process. It assigns a numerical weight (rank) to each webpage based on:

1. The number and quality of links pointing to the page
2. The rank of the pages that link to it

The algorithm uses the following key concepts:

• Sparse Adjacency Matrix: Represents the web as a directed graph where each element represents the probability of moving from one page to another
• Power Method: An iterative approach to compute the final rank values

Features

• Reads network data from a text file
• Constructs a sparse matrix representation of the web graph
• Implements multiple solving methods:
  • Eigenvalue decomposition (for small matrices)
  • Standard Power Method
  • Optimized Power Method
• Visualizes results with a bar plot
• Identifies and displays top-ranked pages

Prerequisites

• MATLAB R2019b or later

Installation

1. Clone this repository:

git clone https://github.com/rasmussvala/TNA010-Page-Ranking.git

2. Navigate to the project directory in MATLAB

Usage

• Run the algorithm by running: PageRank.mlx
• Change data at the top of PageRank.mlx fileID = fopen('Data/data-course-book-1.txt', 'r');