In this week's TidyTuesday video, I go over the feature engineering and modeling process for the UFC sports betting model. I show how to create aggregate features using Dplyr and then explain the different feature components and how it relates to other models. I also bring ideas from past videos such as adding an ELO score to the fight data set. I then explore different ideas for modeling UFC outcomes from a simple binary problem to ensembling multiple components and explain the pros and cons. Next, I show how I modeled each component and outcomes of the fights using XGBoost. Finally, I show how to use different neural network architectures and explain how to solve the complications that come with using multi-headed neural networks.
#DataScience #MachineLearning #TidyModels
Q&A Video Submission Form: https://forms.gle/pjBTJMuDbePvqmug7
Connect with me on LinkedIn: https://www.linkedin.com/in/andrew-couch/
Code for this video: https://github.com/andrew-couch/Tidy-Tuesday/tree/master/Season%201/Apps/TidyTuesdayUFCProject
TidyTuesday: https://github.com/rfordatascience/tidytuesday