Survival Datasets

Survival datasets, containing time-to-event data often with censoring, are crucial for analyzing phenomena like patient survival or equipment failure. Current research focuses on developing accurate and interpretable models, employing diverse architectures such as deep neural networks (including variations like SurvReLU and NSOTree), generalized additive models, and random forests, often incorporating feature selection techniques. These advancements aim to improve prediction accuracy while addressing the challenges of censoring and data heterogeneity, particularly in applications like healthcare where interpretability is paramount for clinical decision-making and trust in model predictions. Federated learning approaches are also emerging to enable collaborative analysis of sensitive survival data across multiple institutions while preserving privacy.