Prediction of Medical Outcomes with Modern Modelling Techniques

The aim of this research is to investigate in what circumstances and under what conditions relatively modern modelling techniques such as support vector machines, neural networks and random forests have advantages in medical prediction research over more classical modelling techniques, such as linear regression, logistic regression and Cox regression.

Specific research questions:
Question 1:
Comparison of modern and traditional modelling techniques:
- What is the performance in predicting intracranial findings on CT scans?
- What is the ability to capture nonlinearity?
Question 2:
Application of modern modelling techniques:
- How can they be applied for survival problems?
- How can they be applied for feature selection in a domain with many variables and comparatively few subjects or data points? Question 3:
Performance of modern modelling techniques:
- What is the performance in relation to the sample size?
- What is the stability of the performance at external validation?