Football Intelligence (FI) is a system for gathering, storing, analyzing, and providing access to data to help Football enthusiasts discover more about the performance of their favorite past time.
While taking Dr. Nelson's Collective Intelligence class I became fascinated with techniques for mining useful data from the "collective intelligence" of readily available data on the Internet.
We decided to apply some of the Data Mining Techniques covered in class in an attempt to predict the 2009 NFL Football season. There is a plethora of data out there that could be mined from Injury reports to betting lines but we decided to limit the scope to use the box score data for training and predictions.
Using box scores from 2003 to present we trained a number of different models from Support Vector Machines to Multilayer Perceptron Networks. The implementations of the models we are using are based on the Weka Data Mining Software. Weka contains a number of tools for experimenting with and visualizing data.
For comparison and to provide some controls we have chosen a few schemes like Home team always wins, City Population, and best Mascot. For the best mascot competition I had my daughters rank the mascots from best to worst and that ranking will be used throughout the season. Poe from the Baltimore Ravens came out on top.
If you would like to see how we are doing or even join us with your own predictions we have pick'em leagues for straight and against the spread.