Comparing classical statistical and machine learning models in survival analysis

Document Type : Original Article

Authors

1 Department of Statistics, Imam Khomeini International University, Qazvin, Iran.

2 Department of Applied Mathematics, Imam Khomeini International University, Qazvin, Iran.

Abstract
Purpose: The purpose of this paper is to evaluate the performance of classical survival analysis techniques and contemporary machine learning algorithms in predicting survival outcomes among patients diagnosed with Heart Failure (HF). The study aims to determine whether machine learning–based models provide improved predictive accuracy compared to conventional statistical approaches.
Methodology: The study applied Kaplan–Meier, log-rank, Cox, and Weibull models alongside machine learning techniques including decision tree, random forest, Inverse Probability of Censoring Weighting (IPCW) logistic regression, Support Vector Machine (SVM), and a survival neural network. Model performances were evaluated based on predictive accuracy.
Findings: The Cox model performed best among classical methods, while IPCW logistic regression and random forest showed the highest accuracy among machine learning models. The survival neural network performed similarly to Cox and random forest without significant improvement.
Originality/Value: The paper provides a concise comparison showing that machine learning models can outperform traditional survival analysis approaches in predicting heart failure outcomes.

Keywords

Subjects

[1]   Miller, J. R. (1981). Survival analysis. John Wiley & Sons. https://books.google.com/books?id=cttQAAAAMAAJ
[2]   Lucijanic, M., & Petrovecki, M. (2012). Analysis of censored data. Biochemia Medica, 22(2), 151–155. https://hrcak.srce.hr/83219
[3]   Deo, S. V., Deo, V., & Sundaram, V. (2021). Survival analysis—Part 2: Cox proportional hazards model. Indian Journal of Thoracic and Cardiovascular Surgery, 37(2), 229–233. https://doi.org/10.1007/s12055-020-01108-7
[4]   Weisberg, S. (2005). Applied linear regression. McGraw-Hill. https://doi.org/10.1002/0471704091.
[5]   Vallarino, D. (2024). Machine learning algorithms for survival analysis: Advantages, disadvantages, and examples. International Journal of Artificial Intelligence and Machine Learning, 4(1), 10–21. https://doi.org/10.51483/IJAIML.4.1.2024.10-21
[6]   Khodaverdi, M. (2018). Survival prediction in heart failure patients using machine learning approaches [Thesis].
[7]   Srujana, B., Verma, D., & Naqvi, S. (2024). Machine learning vs. survival analysis models: A study on right censored heart failure data. Communications in Statistics - Simulation and Computation, 53(4), 1899–1916. https://doi.org/10.1080/03610918.2022.2060510
[8]   Nichols, J. A., Herbert Chan, H. W., & Baker, M. A. B. (2019). Machine learning: Applications of artificial intelligence to imaging and diagnosis. Biophysical Reviews, 11(1), 111–118. https://doi.org/10.1007/s12551-018-0449-9
[9]   Ahmad, T., Munir, A., Bhatti, S. H., Aftab, M., & Raza, M. A. (2017). Survival analysis of heart failure patients: A case study. PloS one, 12(7), e0181001. https://doi.org/10.1371/journal.pone.0181001
[10] Dong, G., Mao, L., Huang, B., Gamalo-Siebers, M., Wang, J., Yu, G., & Hoaglin, D. C. (2020). The inverse-probability-of-censoring weighting (IPCW) adjusted win ratio statistic: An unbiased estimator in the presence of independent censoring. Taylor & Francis, 30(05), 882–899. 10.1080/10543406.2020.1757692
[11] Katzman, J. L., Shaham, U., Cloninger, A., Bates, J., Jiang, T., & Kluger, Y. (2018). DeepSurv: Personalized treatment recommender system using a Cox proportional hazards deep neural network. BMC Medical Research Methodology, 18(1), 24. https://doi.org/10.1186/s12874-018-0482-1
[12] Fotso, S. (2018). Deep neural networks for survival analysis based on a multi-task framework. https://doi.org/10.48550/arXiv.1801.05512