Machine Learning and Graph Algorithms in Building a Very Large Scale Identity Graph
We at Oracle Data Cloud are building an identity graph that connects all the fragmented presence of a person, both online and on the real world, into a single concise view. This allows us to better model a person behavior, target her, and measure the effect of a campaign on her actions.
To build this graph, we apply a tiered model that can handle the very large scale of the graph (composed of billions of points) and be able to utilize the different signals we collect for a person presence.
This model uses a collection of Machine Learning and Graph Algorithms that enable us to solve for identity while controlling the scale and the accuracy of the results.
In this talk I’ll present the challenges of building a large scale probabilistic graph and the set of constrains it imposes on a solution. I would than describe in high level our approach for building this graph and how the handle those challenges.