The text discusses user segmentation in a peer-to-peer payment network using graph-based features. The authors use Neo4j, a graph database, and the Graph Data Science library to analyze the network and segment users based on their roles and positions within the network. They define six features that describe user roles, including average transaction amount, years since first transaction, weighted in-degree, weighted out-degree, betweenness centrality, and closeness centrality. The authors use these features to cluster users into groups or communities using the K-means algorithm, which is a widely used unsupervised machine learning technique. The clustering results show that some users are power users who have sent vast amounts of currency to other users, while others are older accounts with smaller transaction amounts. The visualization of the clusters reveals that the split between clusters isn't very distinct, likely due to the dimensionality reduction algorithm used. Overall, the article demonstrates how graph-based features and machine learning algorithms can be used to segment users in a peer-to-peer payment network.