Predicting Lung Cancer Survival: A Comparative Study of Statistical and Machine Learning Methods

Identifier:  TFM:2527

Handle: https://hdl.handle.net/20.500.11797/TFM2527

Authors:  Correa de Cabo, Ihona Maria

Abstract:
This study explores the application of traditional statistical and Machine Learning models in predicting survival outcomes for lung cancer patients. Using a dataset from “Hospital Universitari Arnau de Vilanova” (2012–2021), it compares Cox and Weibull models with Random Survival Forests, Gradient Boosting, and neural network approaches. Evaluation metrics include the Concordance Index, Integrated Brier Score, and feature importance. Results show that Machine Learning models, particularly Random Survival Forests, slightly outperform traditional methods in predictive accuracy. However, statistical models remain more interpretable. The study highlights the need for improving the interpretability of Machine Learning models to enhance their clinical relevance.