Hybrid GraphSAGE–ANN for Lung Cancer Risk Prediction Using Patient Similarity Graphs

The study contributes a leakage-free and scalable GraphSAGE–ANN hybrid framework for lung cancer risk prediction that effectively integrates patient-level clinical features with graph-based relationships. Its comprehensive comparison with ML, ANN, and multiple GNN models demonstrates the effectiveness of the proposed approach, while strict post-split SMOTE application ensures reliable and fair evaluation.