This project was done as part of EPFL’s CS-433 Machine Learning course in fall 2024.

The project description and instructions can be found here. We were only allowed to use numpy and Python’s standard library for the whole pipeline, hence the manual implementation of all the data cleaning, ML models, selection and validation.

This project aims to use to data from US health surveys to predict the risk of getting coronary heart disease (CHD) or myocardial infarction (MI). We use data from the Behavioral Risk Factor Surveillance System (BRFSS), a system of health-related telephone surveys that collects state data about U.S. residents regarding their health-related risk behaviors, chronic health conditions, and use of preventive services.

You will need numpy, jupyter and matplotlib to run the file, and besides Python’s standard library, there should be nothing else required.

The dataset files should be placed in a folder called dataset next to the run.ipynb file. Run the run.ipynb file to reproduce our results. This includes data cleaning, feature transformation, feature selection, model training, hyperparameter selection, cross-validation, and generation of the predictions on the test set. Some cells will take several minutes to run, this is normal.

• Léo Paoletti
• Thomas Kuntz
• Gabriel Alejandro Jiménez Calles