This project aims to evaluate machine learning methods for healthcare applications, focusing on benchmarking predictive models on a large set of healthcare-related tasks. The first part of the project concerns systematic hyperparameter tuning of classification and regression algorithms. Multiple model classes and hyperparameter configurations will be evaluated across healthcare prediction tasks using appropriate validation procedures and performance metrics. The second part focuses on studying how model performance changes with the amount of available training data. In particular, models will be trained repeatedly using different amounts of data, allowing the contributions of bias and variance to be estimated separately from irreducible noise. This analysis will provide insight into different model classes' behavior in low- vs. high-data regimes, specifically in the healthcare domain. Together, these two parts of the project are aimed at providing a practical benchmark for machine learning within healthcare, as well as a more detailed understanding of the relationship between model choice, hyperparameters, training data, and predictive performance. The computational resources will enable reproducible large-scale experiments involving multiple datasets, model configurations, hyperparameter settings, training-set sizes, and repeated model fits. The project is conducted within the Healthy AI Lab. The main supervisor is Fredrik Johansson, Chalmers University of Technology.