nfl_py3

My first end-to-end machine learning project: Predicting nfl games with data scraped from pro-football-reference.com

To use this:

1. Run build_dataset.py with start_year = 2009
2. Run nfl_cleane.R script to format data correctly
3. nfl_SVM.R will predict the outright winner of a game but, without the spread ~64% accurate
4. nfl_MLP.R will predict the outright winner of a game against the Vegas spread ~56% accurate

The below graph is the accuracy of the MLP on the test set

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 2.0 Generic License.