Optimizing Prediction Accuracy Using Advanced Ensemble And Voting Classifier Methods

This project observes how various machine learning models, once tuned, can further be combined to create a complex model that uses NFL data from the past 18 years to predict the outcomes of matchups between any two competing teams.

Machine learning models have become more advanced over the last few years, allowing for increased prediction accuracy. One relatively undiscovered model-making technique is the Voting Classifier, which is able to combine the predictions of different ML models to arrive at one output prediction. Harnessing this method can allow for increased accuracy and revolutionize the future of Artificial Intelligence (AI) and Machine Learning (ML). This project observes how various machine learning models, once tuned, can further be combined to create a complex model that uses NFL data from the past 18 years to predict the outcomes of matchups between any two competing teams. Through an extensive study of various existing models and key parameters that are critical to the sport of American football, a sophisticated advanced model was created that is highly sensitive to a handful of most important parameters in accurately predicting a given matchup at a level that consistently matches or beats Vegas predictions.

Explore More!

Research Paper

Project Webpage