Machine learning for survival analysis on clinical data

Abstract

The usage of Machine Learning in medicine is a new and a very fast moving technology which is getting more and more attention by information technology companies, doctors, patients and scientists. This technology holds promise for several aspects of medicine, including improving diagnosis of disease, early detection of disease and personalized health care. Currently, experiments with real-world clinical data are necessary to investigate how models based on di erent statistical analysis methods perform in clinical practice. Previous research has observed and measured the influence of various predictors on survival after a cardiac arrest event, both in the form of biomarkers present in the results of blood analysis and from other types of patient information. This master's thesis project continues this study, trying to nd better models using di erent advanced modeling methods for the prediction of several factors related to disease outcome using a large and comperhensive dataset of clinical data.