Application of Diversity-Enhanced Improved Genetic Algorithm to Hyperlink Prediction Problems

Hypergraphs can represent relationships among multiple entities. The link prediction problem, as a classical task in ordinary graphs, has been revisited in hypergraphs and extended to the hyperlink prediction problem. Hyperlink prediction evaluates the likelihood of a node combination serving as a potential hyperedge based on known information in an existing hypergraph. Given a prediction function, this problem faces the challenge of an intractably large solution space that prevents exhaustive search for candidate hyperedges. This paper transforms the prediction function into an optimization objective. By adapting diversity enhancement method to the genetic algorithm, HOGA efficiently identifies suitable candidate hyperedges. Compared to conventional genetic algorithms, it achieves a 32.27% improvement in iteration efficiency with enhanced convergence characteristics.