Statistical link prediction comparison based on the similarity for MEG networks of epilepsy patients
Haeji Lee, Chun Kee Chung, Jaehee Kim · Korean Journal of Applied Statistics · 2025
Link prediction is a network analysis method that predicts the existence and weight of a link between nodes based on a network topology and node characteristics.Link prediction in brain networks plays a vital role in studying the pathological mechanisms of brain diseases and understanding brain network evolution processes.We achieve this by using link modifications, which are changes made to the connections between brain regions, such as structural optimization, finding potential relationships between brain regions, and identifying links.Our study on link prediction methods in brain MEG networks is comprehensive.We explore probability modelbased, similarity index-based, and maximum likelihood estimation-based methods.In addition, we propose new similarity indicators as link predictions based on similarity indices and compare them with existing methods.This thorough approach ensures the validity and reliability of our findings in brain network link prediction in epilepsy patients.