NODES 2025 - Enhancing Benefit Adjudication Through Graph Node Embedding, Clustering, and Outlier Detection

In this presentation, we show how a benefit adjudication process can be improved with graph data science techniques. Our goal is to capture complexities and to reveal patterns and nuances in data sets that would otherwise be difficult using conventional analysis techniques. We demonstrate how graph node embeddings combined with clustering can effectively detect outliers in a realistic patient network. Our program includes modeling hospitalization and benefit data, ingestion into a Neo4j graph database, generating node embeddings, finding similarity among patients, and using similarity to identify outliers. We construct a synthetic graph populated with realistic adjudication data, which we use to evaluate two node embedding algorithms—node2vec and GraphSAGE. These algorithms are chosen because they emphasize different features that need to be incorporated into the two models, allowing us to demonstrate their contrasting strengths and limitations. With the resulting embeddings, we perform clustering analysis to identify outliers in the graph. Speakers: Manish Mithaiwala & Yao Ma Resources: Get Started with Aura - https://bit.ly/3LOLrjh Deployment Center - https://bit.ly/4jOelM3 Ground AI Systems and Agents with Neo4j - https://bit.ly/4oVsnyb #nodes2025 #neo4j #graphdatabase #graphrag #knowledgegraph