Comparison of Some Statistical and Machine Learning Models for Continuous Survival Analysis - research submitted for a Masters degree

Aug 4, 2026

Upload by Riley

While statistical models have long been used for time-to-event problems, machine learning methods are increasingly explored for continuous survival analysis. This comparative study assesses predictive accuracy of statistical and machine learning models using two datasets: time to first alcohol intake and North Carolina recidivism data. Variable selection for both datasets is performed with LassoCV, followed by Kaplan-Meier survival curves and logrank tests. Models compared include CoxPH, Lasso-regularized Cox, Survival Trees, Random Survival Forest, and Neural Networks, evaluated using Integrated Brier score, AUC, and concordance index.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages97

Reading Time4 h 4 min

AccessFree to download

1 Introduction

1.1 Introduction

1.2 Special Features of Survival Analysis

1.3 Discrete and Continuous Time Survival Analysis

1.4 Machine Learning Algorithms

1.5 Problem Statement

1.6 Research Questions

1.7 Research Aim and Objectives

1.8 Outline of the Dissertation

2 Literature Review

2.1 Introduction

2.2 Historical Literature of the Development and Application of the Cox Model

2.3 Historical Literature of the Development of the Discrete Survival Model

2.4 Machine Learning Methods in Survival Analysis

2.5 Comparison of Statistical and Machine Learning Models in Survival Analysis

2.6 Conclusion

3 Research Methodology

3.1 Introduction

3.2 The Survival and Hazard Functions

3.3 Kaplan-Meier Estimator

3.4 Cox PH Model

3.5 Variable Selection

3.6 Machine Learning Algorithms

Comparison of Some Statistical and Machine Learning Models for Continuous Survival Analysis Ndou Sedzani Emanuel (Student No: 18016736) A research submitted in partial fulfilment of the requirements for the degree of Masters of Science in E Science in the Department of Mathematical and Computational Sciences, Faculty of Science Engineering and Agriculture. Supervisor: Dr T.B Mulaudzi Co-supervisor: Dr . A Bere Date Submitted: July 17, 2024 Abstract While statistical models have been traditionally utilized, there is a growing interest in exploring the potential of machine learning techniques. Existing literature shows varying results on their performance which is based on the dateset employed. This study will conduct a comparative evaluation of the predictive accuracy of both statistical and machine learning models for continuous survival analysis utilizing two distinct datasets: time to first alcohol intake and North Carolina recidivism data. LassoCV was used to select variables for both datasets by encouraging limited coefficient estimates. Kaplan-Meier survival curves were utilized to compare the survival distributions among groups of variables incorporated in the model, alongside the logrank test. The proposed methods include the Cox Proportional Hazards, Lasso-regularized Cox, Survival Trees, Random Survival Forest, and Neural Networks. Model performance was evaluated using Integrated Brier score (IBS), Area Under the Curve and Concordance index. Our findings shows consistent dominance of Neural Network (NN) and Random Survival Forest (RSF) models across multiple metrics for both datasets. Specifically, Neural Network demonstrates remarkable performance, closely followed by RSF, CoxPH and CoxLasso models with slightly lower performance, and Survival Tree (ST) consistently lags behind. This study can contribute to advancing knowledge and provides practical guidance for improving survival in recidivism and alcohol intake. Key words: Survival analysis, Statistical models, Machine Learning models, Integrated Brier score, Concordance index, Area Under the Curve. ii Declaration I, Sedzani Emanuel Ndou of Student Number: 18016736, declare that the research titled“Comparison of some statistical and machine learning models for continuous survival analysis”, conducted as part of my Masters of Science in E Science program at the University of Venda, is entirely my own work. I attest that it has not been previously submitted for any degree at any other university or institution. Additionally, I confirm that the content does not include the work of others unless explicitly acknowledged and properly referenced. Student: ......... ................... ................... Date: .1.7..J.u.ly 2024...................................... iii Acknowledgements First and foremost, I would want to express my deepest gratitude and appreciation to God of Mount Zion, who supported me during the whole process by providing blessings, wisdom, and direction, enabling me to finally finish this project. I also extend my appreciation to my supervisor, Dr T.B. Mulaudzi, and my co-supervisor, Dr A. Bere, for their dedicated efforts that contributed greatly to the success of my research project, as well as their support and patience during my research project. Finally, I want to express my gratitude to my family for their support and love. iv Dedication This research project is dedicated to the Republic of South Africa. v Contents Abstract ii Declaration iii Acknowledgement iv Dedication v Abbreviations xiv 1 Introduction 1 1.1 Introduction ................................. 1 1.2 Special Features of Survival Analysis ................... 2 1.3 Discrete and Continuous Time Survival Analysis ............. 3 1.3.1 Discrete time survival analysis ................... 3 1.3.2 Continuous time survival analysis ................. 4 1.4 Machine Learning Algorithms ....................... 5 1

FAQs

Which datasets are used for the comparison in continuous survival analysis?

The study compares models using two datasets: time to first alcohol intake and the North Carolina recidivism data.

How are variables selected before fitting the survival models?

LassoCV is used for variable selection in both datasets by encouraging limited coefficient estimates.

Which evaluation metrics are used to compare the model performances?

Model performance is assessed using Integrated Brier score (IBS), Area Under the Curve (AUC), and the concordance index.

Aug 4, 2026

Upload by Riley