Scalable Hypergraph Learning: Message Passing, Prediction, and Construction

This thesis advances hypergraph learning, a technique that helps computers understand complex group relationships. It introduces three methods that improve how information spreads across hypergraphs, how missing hyperedges are predicted, and how text can be used to build meaningful hypergraphs. These methods make hypergraph models more accurate, scalable, and practical. The research provides new tools for analysing large, interconnected data in areas such as social behaviour, scientific discovery, and intelligent recommendation.