{"id":511,"date":"2019-07-29T20:52:11","date_gmt":"2019-07-29T20:52:11","guid":{"rendered":"http:\/\/172.31.10.168:8081\/blog\/index.php\/2019\/07\/29\/weighted-pagerank-algorithm-in-neo4j-using-cypher\/"},"modified":"2026-01-27T18:18:31","modified_gmt":"2026-01-27T18:18:31","slug":"weighted-pagerank-algorithm-in-neo4j-using-cypher","status":"publish","type":"post","link":"https:\/\/6figr.com\/blog\/weighted-pagerank-algorithm-in-neo4j-using-cypher-511","title":{"rendered":"Weighted PageRank algorithm in Neo4j using Cypher"},"content":{"rendered":"
\"\"<\/figure>\n

PageRank is an algorithm used to measure the relative importance of nodes in a graph system. It is the underlying algorithm that determines what the 3.8 billion population of internet users experience on the web\u00a0today.<\/p>\n

\"\"
The size of each face is proportional to the total size of the other faces which are pointing to it. (source: Wikipedia)<\/figcaption><\/figure>\n

PageRank works great on an unweighted graph, i.e. it assumes that all the edges in the graph are of equal importance. However, there are a number of real world scenarios where the relationship between two nodes may differ based on its\u00a0nature.<\/p>\n

If an article on Medium has links to two other independent web pages B & C, the importance of the link may be inferred from where it is located in the article content ( Higher position => More important => more weight\u00a0).<\/p><\/blockquote>\n

Weighted PageRank<\/strong><\/h4>\n

In such scenarios, we need to resort to an edge-weighted graph and the PageRank algorithm can be modified to incorporate the edge weights in the calculation.<\/p>\n

\"\"
Weighted PageRank calculation. The parameter d is a damping\u00a0factor<\/em><\/figcaption><\/figure>\n

PR_W(p) is the weighted pageRank of P. Here, W(P1.P) denotes the weight of edge from P1 to P, while \u03a3W(P1) is the total weight of all outgoing edges from P1. If this looks overwhelming to you, i\u2019d suggest you read about the basic pagerank calculation here<\/a>.<\/p>\n

Neo4j is one of the most popular graph databases today. Its intuitive modelling and simple query language (cypher) makes it very easy to represent any system as a Graph system. Neo4j does not provide PageRank computation out of the box, hence one needs to use a third party library APOC<\/a> which provides a lot of graph algorithms including PageRank which can be run as cypher procedures in\u00a0Neo4j.<\/p>\n

However, the library does not have support for a Weighted PageRank algorithm. Hence, I decided to implement it in cypher, the code gist of which looks like this\u00a0\u2014<\/p>\n