search for


Simulation study for the machine learning models and evaluation measures in survival data
Journal of the Korean Data & Information Science Society 2023;34:533-65
Published online July 31, 2023;
© 2023 Korean Data and Information Science Society.

Hyunjin Song1 · Jooyi Jung2 · Seungbong Han3

123Department of Biostatistics, Korea University
Correspondence to: This work was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT)[No. 2022R1F1A1063027] and the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI) funded by the Ministry of Health & Welfare, Republic of Korea [No. HI22C045400;No. HA21C0110].
1 Graduate student, Department of Biostatistics, Korea University, Seoul 02841, Korea.
2 Graduate student, Department of Biostatistics, Korea University, Seoul 02841, Korea.
3 Associate professor, Department of Biostatistics, Korea University, Seoul 02841, Korea. E-mail:
Received April 11, 2023; Revised June 19, 2023; Accepted June 24, 2023.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ( which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Better predictions in survival prediction models are an important challenge because it can reduce economic waste and patient pain. Moreover it can increase patient survival through the rapid diagnosis of disease. In reality, the proportional hazard assumption is often violated. In that case, we can use machine learning algorithms which do not depend on the proportionality assumption. Therefore we considered three machine learning models (Cox proportional hazards deep neural network (DeepSurv), Random survival forest (RSF), Survival gradient boosting decision tree (SurvXGBoost)) in this paper. In light of simulation results, we suggest using the Cox proportional hazards deep neural network (DeepSurv) model among the machine learning models. We also claim that it model would be still worthy to consider even in a situation that proportional hazard assumption is somewhat not satisfactory. Meanwhile, it is difficult to propose a specific method for evaluating the performance of survival prediction models. Because the performance measures produce quite different results depending on the choice of hazard function and the censoring rate. Additionally, we compared the results of the model performance based on the breast cancer data set used in Royston and Altman.
Keywords : Machine learning model, performance measure, survival prediction model