Pagerank uses a simplistic model of web surfing to estimate the probability of browsing to each site on the internet.

The model of web surfing works like this:
85% of the time you randomly choose a link on the page you are on and visit that (if there are any links)
15% of the time you choose a random site on the internet and go there
The exact percentages above aren't sacred and can be fiddled with.

The pagerank is the proportion of times you visited each site in this browsing in a very long browsing session.

This works for any kind of graph, and computes a kind of "centrality" measure—how well connected each vertex is to other well connected vertices.

The most common method of computing this is called the power method, and it is essentially just a simulation of this random web surfing recast in matrix/vector terminology. In this algorithm you create a matrix A which contains the transition probabilities described above, so that Aij gives the probability of going from site i to site j. Then you initialize a vector p so all entries are even. You compute the pagerank by repeatedly updating p' = Ap and renormalizing p'. You repeat until p stops changing much from iteration to iteration. The ith element in p is the pagerank of the ith site.

Project Maker Vadodara, Project Maker