Learning and Naming Subgroups with Exceptional Survival Characteristics

2026-02-25

AI Summary

Existing methods for discovering survival subgroups are often constrained by the proportional hazards assumption, the need for pre-discretization of features, and insufficient modeling of individual heterogeneity. This work proposes Sysurv, an end-to-end differentiable, nonparametric framework that, for the first time, integrates random survival forests with differentiable rule learning to automatically construct intrinsically interpretable rules that identify subgroups with significantly deviant survival outcomes relative to the overall population. By eliminating the need for prespecified model assumptions or feature discretization, Sysurv captures individual-level survival deviations in a flexible and data-driven manner. Evaluations on multiple real-world datasets and cancer case studies demonstrate that Sysurv effectively uncovers clinically or engineering-relevant subgroups with high interpretability and actionable insights, thereby validating its efficacy and practical utility.

Technology Category

• Machine Learning: Ensemble Methods
• Search and Optimization: Mixed Discrete/Continuous Search
• Reasoning under Uncertainty: Probabilistic Programming

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

In many applications, it is important to identify subpopulations that survive longer or shorter than the rest of the population. In medicine, for example, it allows determining which patients benefit from treatment, and in predictive maintenance, which components are more likely to fail. Existing methods for discovering subgroups with exceptional survival characteristics require restrictive assumptions about the survival model (e.g. proportional hazards), pre-discretized features, and, as they compare average statistics, tend to overlook individual deviations. In this paper, we propose Sysurv, a fully differentiable, non-parametric method that leverages random survival forests to learn individual survival curves, automatically learns conditions and how to combine these into inherently interpretable rules, so as to select subgroups with exceptional survival characteristics. Empirical evaluation on a wide range of datasets and settings, including a case study on cancer data, shows that Sysurv reveals insightful and actionable survival subgroups.

Problem

• subgroup discovery
• survival analysis
• exceptional survival
• interpretable rules
• individual survival curves

Innovation

• differentiable survival analysis
• non-parametric subgroup discovery
• interpretable rule learning
• random survival forests
• exceptional survival subgroups