Graph Machine Learning in Practice: From Fraud Analysis to GraphRAG
A Talk by Hans Viehmann (Director of Product Management, Oracle)
About this Talk
Combining machine learning techniques with graphs is becoming increasingly popular, not least because of the current advancements in the availability of powerful large language models (LLMs).
Workflows based on Retrieval-Augmented Generation benefit significantly from using graphs in the retrieval step, producing more accurate results while making the results more easily explainable.
That said, even without LLMs, graphs can significantly enhance machine learning models, enabling organizations to uncover hidden patterns and dependencies in data that other methods might miss.
In this talk, we will share the experience from implementation projects, primarily in the financial services industry, where customers use graph pattern matching and different graph-empowered machine learning algorithms for both supervised and unsupervised learning.
We will show how these techniques are used for fraud analytics based on structured data or similarity searches in a GraphRAG scenario with text documents. The presentation will cover how to prepare and model the data, review the methodology, and briefly discuss the results.
We’re a cloud tech company that provides organisations around the world with computing infrastructure and software to help them innovate, unlock efficiencies and become more effective. We also created the world’s first – and only – autonomous database to help organise and secure our customers’ data.
