Machine Learning
Survival Analysis

Latest from the Group

Announcements, releases and events from the group.

19 October 2026 Pinned

Machine Learning for Survival Analysis is available for pre-order

Our textbook with Chapman & Hall/CRC is now listed for pre-order. Andreas Bender and Raphael Sonabend introduce the basic principles of survival and event-history analysis before turning to specific learners, from established estimators to random forests, boosting and neural networks. A particular focus is on the evaluation of survival predictions and on reduction techniques, that is, methods that transform a survival task into a standard regression or classification task — the full text stays free to read online.

Pre-order Read online

10 May 2026 Paper

Reduction Techniques for Survival Analysis

Survival tasks can be turned into ordinary regression or classification problems without ignoring what makes survival data special. The paper surveys the available reductions — piecewise exponential, discrete time, survival stacking, IPCW and pseudo-values — weighs their strengths and weaknesses, and provides implementations that plug into standard machine learning workflows, with a benchmark against established survival methods.

Piller, Orsini, Wiegrebe, Langbein, Burk, Zobolas, Studener, Goeswein & Bender · Lifetime Data Analysis 32(2):34

Paper Code

16 April 2026 Paper

A Large-Scale Neutral Comparison Study of Survival Models

The first large-scale neutral benchmark for single-event, right-censored, low-dimensional survival data: 21 models, from classical statistical approaches to common machine learning methods, on 34 publicly available datasets, with tuning and evaluation handled identically for all of them. The point is guidance that practitioners can act on.

Burk, Zobolas, Bischl, Bender, Wright & Sonabend · Bioinformatics 42(5):btag186

Paper

1 October 2025 Paper

On Training Survival Models with Scoring Rules

How should a survival model — a neural one in particular — be trained? The paper compares training objectives, among them the likelihood, the Brier score and the CRPS, and analyses their properties for survival prediction both theoretically and empirically.

Kopper, Rügamer, Sonabend, Bischl & Bender · ECML PKDD 2025, LNCS 16019, 74–91

Paper

18 February 2025 Paper

High-Dimensional Variable Selection with Competing Events

Penalized Cox regression handles each event type on its own and discards what the causes have in common. Cooperative penalized regression instead fits two cause-specific models in alternation, each one handed the other’s coefficients as prior information, so shared effects are used rather than ignored.

Burk, Bender & Wright · Biometrical Journal 67(1):e70036

Paper

Advancing Survival Analysis

We investigate modern machine learning methods for survival and event-history analysis — bridging statistical rigor with scalable computation.

Book

Machine Learning for Survival Analysis

Andreas Bender and Raphael Sonabend bridge the theory of survival analysis with the methods, workflows and evaluation tools of modern machine learning — from nonparametric estimators and the Cox model to random forests, boosting and neural networks. Chapman & Hall/CRC, 1st edition.

Publisher Read online

Key Research Projects

Selected projects from our group, with code and data openly available.

Open-Source Tools

R packages maintained by our group for applied survival analysis and machine learning.

pammtools

R · CRAN

Piece-wise exponential Additive Mixed Modeling tools — flexible hazard modeling via the mgcv framework.

mlr3proba

R · CRAN

Probabilistic supervised learning for mlr3 — survival analysis, density estimation, and probabilistic regression.

DeepPAMM

Python

Deep piecewise exponential additive mixed models — combining the flexibility of PAMMs with deep learning.

Selected Publications

A selection of recent works from the group.

Get in Touch