Keywords: counting processes, functional data analysis, survival analysis, deep learning
This repository contains the code of the thesis: Performing Survival Analysis via Functional Cox-type Regression and a Machine Learning approach: an application to Heart Failure patients.
The text is available here.
The goal of this thesis was to model the history of patients in the framework of counting processes, to use this information in classical and state-of-the-art survival models and to compare them.
- preprocess : prepare the dataset to fit the stochastic processes representing the history of the patients.
- fit_compensators : build the compensators of the counting processes.
- fpca: summarise compensators through functional principal component analysis scores
- survival_process: build the dataset with fpca scores, fit and compare survival models (Cox, deepHit, DRSA)