Matt GrayCS/ECE 5391/13/2019Predicting Winners of NFL Games with a Neural NetworkFootball has become one of the most popular sports in the United States, and every Sunday and Monday night during the fall and winter NFL teams are broadcasted on television and radio to millions of viewers. Many sports analysts give their picks as to who will win beforethe games start, and in the gambling world there are predictions made as to who will win. Many people place bets on these games and try to use some way of predicting who will win. Unfortunately, many of these predictors have some human bias in them because of personal feelings or thoughts on how a team is doing, who is on the team, etc.I will attempt to predict the outcome of for NFL games using a neural network that reduces the human bias term. I will attempt to do this using a multi-layer perceptron using the back-propagation algorithm. This neural network should prove useful as it can detect patterns and classify problems fairly well. In addition, a multi-layer perceptron handles nonlinear cases. As I have seen from various sources, it is possible to use a neural network to predict the winner of a game. I plan to obtain data from the NFL website as they keep detailed statistics for every game that has been played this year and in the previous few years. In addition, the NFL consists of 32 teams each with a 16 game schedule and then a playoff series in a single season. This creates a vast amount of data throughout the course of a season and this data is available from theNFL website. These statistics pertain to each player on the team and as to what they did in the game. There is more data available than I believe I can actually use well. Therefore, I will need to reduce the features to something that is more useful. I plan to use statistics related to yardage allowed and gained, standard deviations of those averages, and possibly win loss record, to train and predict the outcome of an NFL game. Additionally, to get a better idea of how the dataMatt GrayCS/ECE 5391/13/2019Predicting Winners of NFL Games with a Neural Networkaffects the results, I plan on using various permutations of the data with a K Nearest Neighbor to give me a baseline result that I should at least be obtaining with the multi-layer perceptron.References:http://www.nfl.com for data statisticsNewman, M. E. J., and Park, Juyong; A network-based ranking system for US college Football. Department of Physics and Center for the Study of Complex Systems, University of Michigan, Ann Arbor, MI 48109. arXiv:physics/0505169 v4 31 Oct 2005Purucker, Michael C; Neural network quarterbacking: Who different training methods perform in calling the games. IEEE Potentials, August/September 1996.
View Full Document
Unlocking...