> ML_LIBRARY // LIFELINES_v1.0
lifelines
Cameron Davidson-Pilon — Complete survival analysis and time-to-event modeling library for Python.
classical-mlv0.29.0MITqualified
Model Training
Accelerators:
CPU
Distributed Training:No
Model Inference
Inference Accelerators:
CPU
Deployment Targets:server
What It Does
- +Right, left, and interval-censored data handling for time-to-event outcomes
- +Kaplan-Meier survival curves, Nelson-Aalen cumulative hazard, and log-rank hypothesis tests
- +Cox Proportional Hazards model with time-varying covariates
- +Parametric survival models (Weibull, Exponential, Log-Normal, Log-Logistic)
What It Does Not Do
- -Process high-resolution imagery or audio waveforms
- -Train deep neural networks natively (use PyCox for neural survival analysis)
- -Deploy on edge microcontrollers
>Suitable Work Types
- Customer subscription churn modeling predicting exact time-to-cancellation with censoring
- Predictive maintenance estimating equipment time-to-failure (MTBF)
- Medical clinical trials estimating patient survival probability under different therapies
>Unsuitable Work Types
- Standard binary classification where event time and censoring are ignored (use scikit-learn)
- Real-time millisecond web ad auctions
Data Residency Implications
Runs entirely locally in Python memory. Zero data transmission.
Security Considerations
Permissive MIT license. Safe for proprietary SaaS products.
Operational Profile & Known Limitations
Maturity:mature
Learning Curve:low
Ops Complexity:low
Cost Tier:free-oss
> Known Limitations:
- Cox Proportional Hazards model assumes proportional hazards over time; violations require stratified models or time-varying coefficients.
Associated Incident Patterns (Incidentpedia)
Enforce safeguards and monitoring to guard against these documented real-world failure modes:
> Primary Evidence & Benchmark Citations
lifelines Documentationofficial-docs • >=0.28.0, <=0.29.x
2026-09-25HIGH