SEC Football Predictions with Deep Learning

Inspiration

We were inspired to try an AI related project after team discussion and wanted to incorporate it into something widely known and interesting!

What it does

The neural network reads the history of SEC football game scores from 2008-2015, and predicts the outcomes of the 2016 games (we will be checking 2017 too but we needed a dataset to validate our results).

How we built it

We first designed a method for scraping the relevant data from the internet. Then we processed that data into a usable format and passed into the neural network. The network has 5 total layers, with three hidden layers in the middle.

Challenges we ran into

Tons of them - scraping, cleaning, and processing data through several layers is messy and difficult. Overcoming that was a challenge. The neural network itself was a learning process for the whole team since none of us had experience in implementing them.

Accomplishments that we're proud of

• Our pristine dataset
• The deep neural network
• Our GUI

What we learned

All the things - namely that getting clean usable data is very hard, the basics of deep neural networks, and how connect the pieces between the data, the network, and the gui

Next Steps

Build a more robust dataset for greater accuracy in predictions

Built With

• python
• tensorflow