top of page

Unraveling the Hidden Secrets of Data: When AI Meets Graph

Graphs are pervasive in our surroundings; real-world entities are frequently characterized by their relationships with other entities. A group of objects, along with the links connecting them, can be naturally represented as a graph. If this is kind of abstract, we can imagine the graph in a much concrete way. When we were young in a science class, teachers would give us some atom balls in different sizes and some rods in different lengths to let us create a molecule by randomly combining them. We can see the atom balls as nodes, the rods as edges, and the molecule as a graph. What graph analysis does is to try to extract meaningful insights from the relationships or connections, like whether the molecule we build is reasonable or toxic.

Determining whether the molecule is toxic is just one function of graph analysis. We can do much more than this. For example, we create a social network graph. Each person is a node, and the relationship of every two people is the link (edge) in this graph. If I’m a marketing manager, I’m trying to promote the sales of our new product, perhaps new cell phone, new beverage and even a phone application. What can the graph analysis help? By analyzing the transmission of information and influence between people (nodes) in a social network, it can help to identify and quantify key people and influential ones.

There are other more important uses of graph analysis, such as the cyber security. Nowadays, we’ve moved beyond just using phones to make calls. We save our money in electric wallet and our private identity information. Ensuring the safety of the online environment is essential. If I’m a security engineer in a bank, I want to ensure the fund safety of every depositor. By building user behavior and identity association graphs, we can model and analyze user behaviors for authentication and access control to avoid theft and abnormal behavior.

If we connect AI with xGT, a high-performance graph analysis platform, we can combine the data analysis based huge amounts of data with graph algorithms to explore deep relationships. It’s hard to do graph analysis on a large dataset on your local server. It’s not only time consuming, but very complex work. Using xGT, with just couple rows of simple code, you can finish the analysis in real time. With a simple function, we can import the data from xGT to a PyG environment to do some complex model training and prediction. We can then transfer the prediction result back to xGT to do further analysis.

This is just a modest spur to induce future more creative thinking. What are you waiting for? Accelerate and enhance your AI/ML efforts today with Trovares xGT.

Author: Jinyi Zhang

Edited by: Erik Rottsolk


bottom of page