DevMesh

Mining Associations of Active Nodes in Large Graphs

Srivignessh Pss

Srivignessh Pss

6 0
  • 0 Collaborators

Graphs represent our social networks in social media applications, navigation within a website, and any other data which focuses on the relationship between items. Understanding the association between elements is crucial to the development of algorithms which use these relationships to make artificially intelligent decisions. These algorithms recommend new connections on social media, suggest new products for purchase by a customer, serve relevant ads to users, and many other personalization strategies characteristic of modern software products. ...learn more

Project status: Published/In Market

Internet of Things, Artificial Intelligence

Intel Technologies
Other

Code Samples [1]

Overview / Usage

A large graph of 10,000 nodes like Amazon product co-purchasing network is chosen. Its randomly marked with a few hundred active products currently purchased by customers. Our goal is to find the relationships between these active products from the historical relationship from product co-purchasing network to identify the products which are well connected and form groups such that they become the best product recommendations.

Methodology / Approach

In a large graph in which some of the nodes are marked as active based on their recent activity.
We would like to find the relationship between these active nodes and to
determine whether they are close to each other in the graph or they are
forming separate groups. These relationships chosen must be optimal,
for which, we need to segregate these active nodes from the rest and
group them using partitioning algorithms.

The Graph Mining is done by building a dynamic minimum spanning tree of active nodes.
The algorithm is improved by using optimal computation.

Technologies Used

Intel Optimized Distribution of Python is used for building this efficient graph algorithm which saves the runtime from a day to few hours. Neo4j Graph Database is used for storing and retrieving graph data and visualizing them.

Repository

https://link.springer.com/chapter/10.1007/978-3-319-96136-1_4

Comments (0)