Reproducible analysis of limited datasets, project 1 (IN-STK5000 exam project)

This project is an exam project for the course IN-STK5000 at the University of Oslo: "Responsible Data Science" (formerly "Adaptive methods for data-based decision making").

The project is tasked with training an ML classifier to predict whether a patient should be checked for diabetes, using a very limited dataset of 500 individuals with limited data quality. The focus was on implementing a reproducible approach that made it simple to re-run, adapt and evaluate experiments, also for people outside the group. Our assigment was graded to an A, 40 of 40 marks.

Repo for code: https://github.com/YSLCoat/IN-STK5000...

Group members:
Espen H. Kristensen:   / espenhk  
Torstein Forseth:   / torstein-forseth-3a5b68210  
Cornelius Bencsik:   / cornelius-bencsik-4b8096149  
Amir Basic:   / amir-basic