ncaa_ml is a fun project that uses KenPom data to generate March Madness brackets using GNB, RF models.
- the main
ncaa_ml.pyruns a specified number of simulations. - choose one to fill out your bracket with.
- can visualize the end of simulation by who won the most.
- Install requirements.txt with conda / favorite env manager.
- gets all kenpom data and updates config
python get_data.py - if necessary change ncaa_config.yml for # simulations, risk
- in config replace all State -> St. to match names, and fix names that might be wrong like UConn -> Conneticut
- in kenpom_current DELETE the teams that aren't in from first 4, only should have 64 teams
- python ncaa_ml.py --data data/kenpom_data.csv --plot False
The data is scraped from KenPom from 2002-Current Year.
Out of respect for KenPom, I prefer not to supply the scraping script (however it is pretty easy), and the formatted data.
It looks something like this:
